JPH1125218A - Methods for line segmentation and line segmentation modification and devices for line segmentation and line segmentation modification - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain line segmentation without preliminarily or individually judging a line direction even when the line direction of a document picture is not clear. SOLUTION: When a line direction is not clear, or when the line direction is a first direction, a line segmenting device 10 produces a first direction divided area, generates a first direction projective distribution by scanning the above area, and segments a first divided line, based on this distribution. Meanwhile, when the line direction is not clear, or when the line direction is a second direction, this line segmenting device 10 produces a second direction divided area, generates a second direction projective distribution by scanning the above area, and segments a second divided line, based on this distribution. Then, when the line direction is not clear, the line direction is judged by using the information of the first and second divided lines, and the divided lines in the judged line direction are connected with each other so that a line can be segmented.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention optically scans characters on a document to obtain a document image (image data) represented by, for example, a binary image, and performs character recognition on the document image. The present invention relates to a method and an apparatus for cutting out a line of a document image in an optical character reader (OCR), and a method and an apparatus for correcting a line cutout.

[0002]

2. Description of the Related Art In character recognition, generally, first, a document is read by a scanner and lines of a generated document image are cut out. Next, characters are extracted from the extracted lines.
Then, the cut-out character is recognized.

An example of a conventional line segmentation method is disclosed in Reference 1:
It is disclosed in "JP-A-64-48186".
According to the technique disclosed in this document, when performing line segmentation, first, a document image is divided by a dividing line that crosses a character string (line). Next, a projection distribution (projection) in the character string direction is created for each divided area. Next, a partial character string is cut out for each divided area based on the projection distribution.
Next, the same label is assigned to each of the partial character strings in which the cutout coordinates of the partial character strings of the adjacent divided areas overlap. Then, the partial character strings having the same label in the adjacent divided areas are connected to each other, and the line of the document image is cut out.

[0004]

However, the conventional line segmentation method is based on the premise that the line direction of the document image is known. Therefore, in the conventional line extraction method, line extraction of a document image whose line direction is unknown cannot be performed. For this reason, conventionally, when the line direction of a document image is unknown, it has been necessary to determine the line direction individually before performing line segmentation.

[0005] For this reason, there has been a demand for a line cutting method and apparatus capable of line cutting without individually determining the line direction in advance even when the line direction of the document image is unknown.

Further, in the conventional line segmentation method, for example, a case where adjacent lines are close to each other or a line direction of a document image (a character string is When the input is performed with the inclination of the direction in which the constituent characters are arranged), two adjacent individual lines may be cut out as one line.

In the conventional line segmentation method, for example, when each character constituting a certain character string is composed of radicals divided vertically, one line is segmented as two separate lines. There was sometimes.

[0008] For this reason, there has been a demand for a method and apparatus for correcting a line cutout that can ensure the line cutout.

[0009]

[Means for Solving the Problems]

(Line extraction method) The inventor of the present application has made various studies and experiments, and as a result, if the line direction is determined using the information obtained by the line extraction processing, the line extraction is performed before the line extraction. The inventor has conceived that it is possible to cut out a line of a document image whose line direction is unknown without performing a line direction determination process separately from the process.

Therefore, according to the line extracting method of the present invention, (a1) when the line direction of the document image is unknown and when the line direction of the document image is the first direction, the line is orthogonal to the first direction. (B1) generating a plurality of first direction divided regions arranged along the first direction by dividing the document image by dividing lines along the second direction;
When the direction division area is generated, each of the first direction division areas is scanned in the first direction to detect the cumulative number of black pixels for each scan, and is detected in the second direction of the cumulative number of black pixels. (C1) when the first-direction projection distribution is created, the cumulative black that is equal to or more than the first reference value in each of the first-direction projection distributions. A process in which a section in which the number of pixels is continuous along the second direction is a first black pixel section, and a scanning area in which the first black pixel section is obtained is a first division row, and a divided row is cut out; When the line direction of the document image is unknown and when the line direction of the document image is the second direction, the document image is divided by a dividing line along the first direction to be orthogonal to the first direction. Lined up along the second direction (B2) when the second direction divided regions are generated, each of the second direction divided regions is scanned along the second direction to generate a plurality of second direction divided regions. A process of detecting a cumulative number of black pixels and obtaining a distribution of the cumulative number of black pixels along the first direction to form a second direction projection distribution; and (c2) a case where the second direction projection distribution is generated. In each of the second direction projection distributions, a section in which the cumulative number of black pixels equal to or greater than the second reference value is continuous along the first direction is defined as a second black pixel section,
The scanning area in which the second black pixel section is obtained is defined as a second
(D) a process of extracting a divided line as a divided line;
A process of determining the line direction of the document image based on the first direction projection distribution and the second direction projection distribution when the line direction of the document image is unknown; In the case of the direction, the divided rows included in the first direction divided areas adjacent to each other in the first direction,
A process of connecting the divided lines in which at least a part of the first black pixel section obtained by scanning the divided line overlaps to perform line cutout, and (e2) the line direction of the document image is in the second direction. In some cases, the divided rows included in the second direction divided areas adjacent to each other in the second direction overlap each other, and at least a part of the second black pixel sections obtained by scanning the divided rows overlap. A process of connecting the divided rows to perform line segmentation.

As described above, according to the line extracting method of the present invention, when the line direction of the document image is unknown, the first direction projection obtained in the process (b1) which is a part of the line extracting process. Using the distribution and the second direction projection distribution obtained in the process (b2), the row direction is determined in the process (d). Therefore, according to the present invention, even when the line direction of a document image is unknown, line extraction can be performed without performing a process of determining the line direction separately from line extraction in advance.

In order to determine the black pixel section corresponding to each row in the projection distribution obtained by scanning the document image along the row direction, the reference value of the cumulative number of black pixels (for example, the first A reference value and a second reference value are set, and a portion of the accumulated number of black pixels equal to or larger than the reference value is preferably set as a black pixel section. For example, when the document image portion other than the line (character string) in the document image is scanned and the number of accumulated black pixels of one or more pixels is detected (for example, the scanning region of the document image includes a frame line) In this case, if any suitable reference value is appropriately set, the possibility that the black pixel section corresponding to each row is obtained is increased.

Further, for example, when the row direction of a document image is inclined with respect to the scanning direction, the projections of adjacent divided rows may partially overlap in the projection distribution. . In this case, the number of cumulative black pixels in the projection of the overlapping portion is often smaller than the number of cumulative black pixels in the non-overlapping portion. Therefore, any suitable reference value (for example, a numerical value about the cumulative number of black pixels in the overlapping portion, for example, a numerical value between a numerical value about the cumulative number of black pixels in the overlapping portion and a numerical value about the cumulative number of black pixels in the non-overlapping portion) Is set, there is a high possibility that a black pixel section corresponding to each row is obtained.

(Determination of Row Direction Based on Variation in Black Pixel Section Length) By the way, in performing the determination of the row direction in the above process (d), the inventor of the present application has one of the clues of the row direction determination. We focused on variations in row height and row width.

In general, when a document image is written horizontally, the heights of the characters on each line are uniform. When the document image is written vertically, the widths of the characters on each line are uniform. Therefore, in the projection distribution obtained by scanning the document image along the row direction, the variation in the length of the black pixel section (black pixel section length), which is the projection portion corresponding to each row, is generally small. On the other hand, in a projection distribution obtained by scanning along a direction orthogonal to the row direction, generally, there is a large variation in the length of each black pixel section. Therefore, focusing on the magnitude of the variation in the black pixel section length, the row direction can be determined.

Therefore, in carrying out the line extracting method of the present invention, preferably, in the process (d), the first
The process of calculating the variation of the first black pixel section length, with the length of each black pixel section as the first black pixel section length, and the length of each of the second black pixel sections as the second black pixel section As the length, the process of calculating the variation of the second black pixel section length, and comparing the variation of the first black pixel section length with the variation of the second black pixel section length, to determine the first black pixel section length Determining the scanning direction of the projection distribution from which the smaller black pixel section length is obtained as the row direction when the ratio of the variation of the second black pixel section length to the variation of the second black pixel section length is equal to or greater than the third reference value. It is good to include

As described above, the variation in the black pixel section length of the projection distribution obtained by scanning the document image substantially in the row direction is caused by the variation of the projection distribution obtained by scanning the document image across the row. Generally smaller than the variation of the black pixel section length.
Therefore, by comparing the variation in the first black pixel section length with the variation in the second black pixel section length, the determination in the row direction can be assured.

The third reference value may be set to any suitable value. For example, when the third reference value is set to “1”, it corresponds to simply comparing the variation between the first black pixel section and the variation between the second black pixel sections.
The reason why the third reference value is not limited to “1” is that when the ratio of the variation is small, if the row direction is determined only by the magnitude of the variation, an erroneous determination result may be output. is there.

(Judgment of Line Direction Based on Average White Pixel Section Length) The inventor of the present application has paid attention to the line interval and the character interval as one of the clues for determining the line direction.
In a general document image, the line interval is wider than the character interval. This line interval is reflected in the interval between adjacent black pixel sections corresponding to each line in the projection distribution obtained by scanning the document image along the line direction. The character spacing is reflected in the spacing between adjacent black pixel sections in a projection distribution obtained by scanning along a direction orthogonal to the row direction. Therefore, if attention is paid to the line spacing and the character spacing, it is possible to determine the line direction.

Therefore, in carrying out the line segmentation method of the present invention, more preferably, in the process (d),
When the ratio of the variation in the first black pixel section length to the variation in the second black pixel section length is less than the third reference value, the first interval between the adjacent first black pixel sections in the first direction projection distribution. It is preferable to include a process of determining the row direction using the length of the white pixel section and the length of the second white pixel section between the adjacent second black pixel sections in the second direction projection distribution.

By the way, even if the first and second black pixel section length variations are not compared, the line direction can be determined by focusing on the line interval and the character interval.

Therefore, in implementing the line segmentation method of the present invention, in the process (d), the length of the first white pixel section between the first black pixel sections adjacent to each other in the first direction projection distribution; It is preferable to include a process of determining the row direction by using the length of the second white pixel section between adjacent second black pixel sections in the second direction projection distribution.

Further, the first white pixel section length and the second
In performing the row direction determination focusing on the white pixel section length,
More preferably, when there is one first white pixel section,
The length of the first white pixel section is defined as a first average white pixel section length, and when there are a plurality of first white pixel sections, the arithmetic average of the lengths of the first white pixel sections is calculated. The process of calculating as the first average white pixel section length, and when there is one second white pixel section, the length of the second white pixel section is set to the second average white pixel section length, and the second white pixel section When there are a plurality of pixel sections, the process of calculating the arithmetic mean of the lengths of the second white pixel sections as the second average white pixel section length, the first average white pixel section length and the second average white pixel section It is preferable to include a process of comparing with the pixel section length and determining the scanning direction of the obtained projection distribution having the larger average white pixel section length as the row direction.

As mentioned above, line spacing is generally wider than character spacing. Therefore, of the first average white pixel section length and the second average white pixel section length, the larger average white pixel section length corresponds to a blank between rows. Therefore, if the row direction is determined by comparing the first average white pixel section length with the second average white pixel section length, the ratio of the variation of the first black pixel section length to the variation of the second black pixel section length is determined. Even in a small case (for example, when the character interval is a fixed pitch), it is possible to ensure the determination of the line direction.

(Judgment of Line Direction in Case of Only One Line) By the way, when the character string is included in the document image only in one line, each divided area generated by dividing by a dividing line orthogonal to the line direction is generated. Contains one divided line at a time. As a result, in this case, the number of divided areas matches the total number of divided rows included in each divided area. On the other hand, when the document image is divided by the dividing line along the line direction, the number of divided regions does not match the total number of divided lines. For example, even if a plurality of divided areas are generated, the total number of divided rows remains one. Therefore, when the character string is only one line, if the number of divided areas and the total number of divided lines are compared, it is needless to compare the variation of the black pixel section length and the average white pixel section length. , The row direction can be determined.

Therefore, in carrying out the line segmentation method of the present invention, preferably, in the process (d),
A process of calculating the total number of the first direction divided regions as the first divided region number, and a process of calculating the total number of the first black pixel regions over all the first direction divided regions as the first black pixel region number. , The sum of the number of divided regions in the second direction
The sum of the process of obtaining the number of divided areas and the number of second black pixel sections over all the second direction divided areas is defined as a second
If the process for obtaining the number of black pixel sections is equal to the number of first black pixel sections and the number of the first divided areas and the number of second black pixel sections is not equal to the number of the second divided areas, the row direction Is determined as the first direction, and when the number of second black pixel sections and the number of the second divided areas match, and the number of first black pixel sections and the number of the first divided areas do not match. And a process of determining the row direction as the second direction.

As described above, when the number of divided regions and the total number of divided lines are compared, when the character string is only one line, the variation in the black pixel section length and the average white pixel section length are determined. The row direction can be determined without comparison.

(Connection of Divided Rows) By the way, after each divided row is cut out in each divided area, when each divided row is connected, there is a case where one divided row can be connected to two divided rows on only one side. is there. That is, the black pixel section of one divided row may partially overlap the black pixel section of two divided rows on one side only. Such duplication tends to occur when the row direction of the document image is inclined with respect to the scanning direction.

In the case where the row direction is inclined as described above, for example, even in the technique disclosed in the above-mentioned document 1, the cut-out coordinates of the partial character string (corresponding to the divided line) are adjacent to the divided area. Since the cutout coordinates of the two partial character strings overlap with each other, the same label may need to be attached to both of the two partial character strings. In this case, in the technique disclosed in Document 1, line segmentation becomes difficult.

Therefore, in carrying out the line segmentation method of the present invention, preferably, in the process (e1), that is, in the process of connecting the first divided lines, the first direction divided lines included in the divided region are separated. When a part of the first black pixel section overlaps with a part of the first black pixel section of two first divided rows included in another divided area adjacent to the divided area, respectively, Connecting the divided lines whose length is longer than the fourth reference value,
In the process (e2), that is, the process of connecting the second divided rows, the second
When a part of the second black pixel section of the direction divided row overlaps with a part of the second black pixel section of two second divided rows included in another divided area adjacent to the divided area, respectively, the overlap occurs. It is preferable to connect divided rows whose section length is longer than the fifth reference value.

As described above, if the divided rows to be connected are determined using the fourth reference value or the fifth reference value, even if two divided rows overlap on one side of the divided row, the divided rows can be appropriately connected. Can be connected. As a result, line segmentation can be performed by appropriately connecting the divided lines over each divided region.

The fourth reference value and the fifth reference value may be set to any suitable values. For example, the fourth reference value and the fifth reference value may be equal to each other, or may be different from each other. Further, it is preferable that the fourth and fifth reference values are set so that the amount of deviation between the divided rows connected to each other is minimized. For example, when the black pixel section length and the white pixel section length are almost the same length, at least one of the fourth and fifth reference values may be set to a value of about “0.5”. In this case, if one of the two divided rows of the adjacent divided area that overlaps on one side of one divided row satisfies the reference value, the other does not satisfy the reference value.

(Integration Correction 1) By the way, when a divided line is cut out, there is a case where one divided line is originally cut out as two divided lines by mistake. For example, when each character constituting a certain horizontal writing character string is composed of radicals that are vertically divided, one divided line may be originally cut out as two individual divided lines. Also, for example, when each character constituting a certain vertically written character string is composed of radicals that are divided into left and right, one divided line may be originally cut out as two individual divided lines.

The two divided lines erroneously cut out are:
The height (in the case of horizontal writing) or width (in the case of vertical writing) of each divided line is shorter than other divided lines that are appropriately cut out. The height or width of two erroneously divided rows is generally about half that of the other divided rows. Furthermore, the line interval between the two erroneously extracted divided lines is generally smaller than the line interval between the other divided lines.

Therefore, if attention is paid to the height or width and the line spacing of the cut-out divided lines, a divided line that is supposed to be cut out as one divided line is erroneously found as two divided lines. As a result, it is possible to correct the division line cutout.

Therefore, in carrying out the line segmentation method of the present invention, preferably, before performing the process (e1), the length of the first black pixel section of the first divided row segmented in the process (c1) (Hereinafter, also referred to as a first black pixel section length) as a first average black pixel section length, and a process of obtaining the first average black pixel section length and the first average black pixel section length.
By comparing the black pixel section length with the first black pixel section, the first black pixel section in which the ratio of the first black pixel section length to the first average black pixel section length is smaller than the sixth reference value is defined as the first black pixel section. The process of selecting as a candidate section, the first candidate black pixel section which is the nearest first black pixel section on one side of the first candidate section, and the first candidate section as one continuous black pixel section In this case, the black pixel section is defined as a first integrated section, and the process of obtaining the length of the first integrated section, and the magnitude of the difference between the length of the first integrated section and the first average black pixel section length is defined as the first integrated section. A process of obtaining the integrated section length error, a second candidate black pixel section that is the nearest first black pixel section on the other side of the first candidate section, and the first candidate section as one continuous black pixel section In this case, the black pixel section is defined as a second integrated section, and the black pixel section of the second integrated section is A second integrated section length error, a process of obtaining the magnitude of the difference between the length of the second integrated section and the first average black pixel section length as a second integrated section length error, A process of comparing a long error with the integrated section having a smaller error as a first selected integrated section; and a first candidate section and a nearest one of the first candidate section of the first selected integrated section. A process of obtaining a length of a first white pixel section between the first black pixel section and the first black pixel section;
By comparing the length of the white pixel section with the seventh reference value,
When the length of the white pixel section is shorter than the seventh reference value,
It is preferable to include a process of correcting the line segmentation by setting the first selection integrated section as one continuous first black pixel section.

Here, in correcting the line segmentation, first, focusing on the black pixel segment length, a first candidate segment having a high possibility of being erroneously divided and segmented is selected. In selecting the first candidate section, the first black pixel section length is compared with the first average black pixel section length. Then, a first black pixel section whose ratio to the first average black pixel section length is smaller than the sixth reference value is selected as a first candidate section. As the sixth reference value, for example, any suitable value (for example, about 0.5) smaller than 1 and larger than 0 may be set.

Except for the case where the first candidate section is an end row, divided rows exist on both sides of the first candidate section. Then, which of the two divided rows on both sides is to be integrated with the first candidate section to determine which should be corrected. In making this determination, the first integrated section length error and the second integrated section length error are compared, and the one with the smaller error is selected as the first selected integrated section.

Further, when one row is incorrectly divided into two rows, the length of the white pixel section between the two divided rows is generally short. Therefore, the length of the first white pixel section between the first candidate section and the first selection integration section is compared with a seventh reference value. The seventh reference value may be set to any suitable value. For example, a value smaller than the average value of the length of the white pixel section (for example, about half the average value) may be set as the seventh reference value.

When the length of the first white pixel section is shorter than the seventh reference value, the line extraction is corrected with the first selected integrated section as one line. If the length of the first white pixel section is equal to or longer than the seventh reference value, it is highly likely that the first selection integrated section is not originally one line. No correction is made to the line segmentation of the integrated section.

(Integration Correction 2) In carrying out the line segmentation method of the present invention, preferably, before performing the process (e2), the second segment of the second divided line segmented in the process (c2) is performed. Length of black pixel section (hereinafter referred to as second
Also referred to as black pixel section length. ) Is calculated as the second average black pixel section length, and the second average black pixel section length is compared with the second black pixel section length to determine the second average black pixel section length of the second black pixel section. The second black pixel section in which the ratio of the second black pixel section length to the pixel section length is smaller than the eighth reference value is defined as the second black pixel section.
The process of selecting as a candidate section, the third candidate black pixel section which is the second black pixel section nearest to one side of the second candidate section, and the second candidate section are regarded as one continuous black pixel section. In this case, the black pixel section is defined as a third integrated section, and the length of the third integrated section is determined. The difference between the length of the third integrated section and the second average black pixel section length is determined as the third integrated section. The process for obtaining the integrated section length error and the second
A fourth black pixel section, which is the nearest second black pixel section on the other side of the candidate section,
When the candidate black pixel section and the second candidate section are defined as one continuous black pixel section, the black pixel section is defined as a fourth integrated section, and the process of obtaining the length of the fourth integrated section is performed. A process of determining the magnitude of the difference between the length of the integrated section and the second average black pixel section length as a fourth integrated section length error;
A process of comparing the third integrated section length error with the fourth integrated section length error and selecting an integrated section having a smaller error as a second selected integrated section; and a process of selecting a second integrated section of the second selected integrated section. The process of determining the length of the second white pixel section between the candidate section and the second black pixel section immediately adjacent to the second candidate section, and comparing the length of the second white pixel section with the ninth reference value When the length of the second white pixel section is shorter than the ninth reference value, the process of correcting the line segmentation by setting the second selection integrated section to one continuous second black pixel section It is good to include

Also in this case, similarly to the above-described correction of the line cutout for the first black pixel section, the second selected integrated section is selected and the line cutout is corrected. The eighth reference value is
As with the sixth reference value, any suitable value may be set. Also, the ninth reference value may be set to any suitable value, similarly to the seventh reference value.

(Division Correction 1) By the way, at the time of division line extraction, two individual lines may be erroneously extracted as one line. For example, when the adjacent lines are close to each other, or when the line direction of the document image (the direction in which the characters forming the character string are arranged) is inclined with respect to the scanning direction when creating the projection distribution, for example. When input, two adjacent individual lines may be erroneously cut out as one line.

The divided line erroneously cut out as one line is
The height (in the case of horizontal writing) or width (in the case of vertical writing) of the divided line is longer than other divided lines that are appropriately cut out.
In general, the height or width of a divided row that is erroneously cut out as one row is about twice as large as other divided rows.

Therefore, if attention is paid to the height or width of the cut-out divided lines, a divided line that is supposed to be cut out individually as two divided lines is mistakenly found as one divided line. It is possible to correct the split line segmentation.

Therefore, in carrying out the line extracting method of the present invention, it is preferable that the length of the first black pixel section of the first divided row cut out in the process (c1) (hereinafter referred to as the first black pixel section length) The arithmetic mean of the first
The process of obtaining the average black pixel section length is compared with the first average black pixel section length and the first black pixel section length, and among the first black pixel sections, the first black pixel corresponding to the first average black pixel section length A process of selecting a first black pixel section whose section length ratio is larger than the eleventh reference value as a first division candidate section; and a step of selecting black pixels on a scan line that scans the vicinity of the center of the first division candidate section among scan lines. A process for determining, as the first position, a position closest to the first direction along the first direction among the detected positions, and a process for determining the first position closest to the other side along the first direction among the first positions. A process of obtaining one position, and a position closest to the other side in the second direction among positions where black pixels are detected on a scanning line that scans the vicinity of the center of the first division candidate section in the scanning line is a second position. And the second position The process of determining the closest second position along the second direction as the closest second position, and comparing the closest first position and the closest second position, determine that the closest first position is the closest second position. When the second position is on the other side than the second position or when the first position and the second position are located on the same straight line along the second direction, the first division candidate section is
The process of correcting the line segmentation by dividing the first divided candidate section into two at the center of the first divided candidate section to generate two divided black pixel sections, and comparing the nearest first position with the nearest second position Then, when the nearest first position is located on one side of the nearest second position, the nearest first position and the nearest second position are set.
The first line is a dividing line passing through the middle of the position and along the second direction.
By dividing the division candidate section, a first divided area on one side and another divided area on the other side are generated, and one divided area is divided by a scanning line from which a nearest second position is obtained, thereby dividing the divided area into two divided areas. Generating a black pixel section, and dividing the other divided area by the scanning line at which the nearest first position is obtained to generate two divided black pixel sections to correct the line segmentation. good.

Here, in correcting the line segmentation, first, focusing on the first black pixel segment length, a first division candidate segment having a high possibility of being erroneously integrated and segmented is selected.
In selecting the first division candidate section, the first black pixel section length is compared with the first average black pixel section length. And the first
A first black pixel section whose ratio to the average black pixel section length is larger than the eleventh reference value is selected as a first division candidate section. As the eleventh reference value, for example, one or more arbitrary suitable value (for example, about 1.5 to 2) may be set.

If two individual divided lines are erroneously cut out as one divided line, for example,
There are a case where two individual divided lines are close to each other, and a case where the line direction of the document image is inclined with respect to the scanning direction. In either case, the projection distributions of the two individual divided rows may overlap each other and be erroneously cut out as one continuous black pixel section.

For example, when the above-mentioned divided rows are close to each other, the first divided line is formed by a dividing line along the scanning direction.
What is necessary is just to forcibly divide the division candidate section into two. On the other hand, for example, in the case where the row direction is inclined with respect to the scanning travel, it may not be possible to properly cut out each row if the forcible division is performed. That is, if the division is forcibly performed, a part of each of the two rows may be included diagonally in each of the divided rows.

Therefore, here, the nearest first position and the nearest second position are compared. Then, when the nearest first position is one side of the nearest second position, or when the nearest first position and the nearest second position are located on the same straight line along the second direction. (These cases correspond to, for example, the case described above.), The first segmentation section is forcibly divided into two to correct the line segmentation.

On the other hand, when the nearest first position is on the other side of the nearest second position (this case corresponds to, for example, the above case), the first division candidate section is divided into four. By doing so, the line segmentation is corrected.

(Division Correction 2) In carrying out the line extraction method of the present invention, preferably, the length of the second black pixel section (hereinafter, referred to as the second black pixel section) of the second division line extracted in the process (c2) is used. A process of calculating the arithmetic mean of the two black pixel section lengths as a second average black pixel section length;
The second average black pixel section length is compared with the second average black pixel section length, and the ratio of the second average black pixel section length to the second average black pixel section length in the second black pixel section is greater than the twelfth reference value. Selecting the second black pixel section that is also large as the second division candidate section, and selecting the black pixel detection position along the second direction among the positions where black pixels are detected on the scanning line that scans the vicinity of the center of the second division candidate section. A process of obtaining a position closer to one side as a third position, a process of obtaining a third position closest to the other side along the second direction among the third positions as a closest third position, and a second division of scanning lines. A process for determining, as a fourth position, a position closest to the other side along the first direction among positions where black pixels are detected on a scanning line that scans the vicinity of the center of the candidate section, and a first direction among the fourth positions. 4th position closest to you along 4th And processes determined as in the case by comparing the nearest third position and nearest the fourth position, the nearest third position is the other side from the nearest fourth position, the second
The process of correcting the line segmentation by dividing the division candidate section into two in the center of the second division candidate section to generate two divided black pixel sections, and the third position and the nearest third position Compared with the four positions, when the third closest position is one side of the fourth closest position, and when the third closest position and the fourth closest position are along the first direction. When located on the same straight line, the second division candidate section is divided by a division line that passes through the middle between the nearest third position and the nearest fourth position and extends in the first direction. A first divided region and the other divided region on the other side, and the one divided region is divided by a scanning line at which the nearest fourth position is obtained to generate two divided black pixel sections; The area is divided by the scanning line from which the nearest third position is obtained, and the two divided black pixel sections are divided. It may include a process of correcting the line cut by forming.

Here, the second division candidate section is selected in the same manner as in the case of selecting the first division candidate section.
Note that the twelfth reference value may be set to any suitable value as in the case of the eleventh reference value.

Then, in the same manner as the comparison between the closest first position and the closest second position, the closest third position and the closest fourth position are compared, and the first division candidate is determined. Similarly to the case of the section, the second division candidate section is divided into two or four.

(First Line Extraction Correction Method) In the above (Integration Correction 1) and (Integration Correction 2) in the line extraction method of the present invention, the correction of the division line extraction was performed. The method is not limited to the modification of the extraction of the divided line, but can be applied to the modification of the extraction of the line which is not divided.

Therefore, according to the first line segmentation correcting method of the present invention, the document image is scanned along the first direction substantially along the line direction of the document image, and the cumulative number of black pixels for each scan is calculated. A projection distribution is created by detecting the distribution of the cumulative number of black pixels along a second direction orthogonal to the first direction to create a projection distribution. After performing line segmentation with a scanning region in which each black pixel segment continuous in the direction is obtained as one row region, when performing line segmentation correction, the length of the black pixel segment of the segmented row region (black pixel segment) The process of calculating the arithmetic mean of the section lengths as the average black pixel section length, and comparing the average value with the black pixel section length, determine that the ratio of the black pixel section length to the average value among the black pixel sections is based on the 14th standard A black pixel section smaller than the value is defined as a candidate section. And the first candidate black pixel section, which is the nearest black pixel section on one side of the candidate section, and the black pixel section when the candidate section is regarded as one continuous black pixel section. One integrated section and the first
A process of obtaining the length of the integrated section, a process of obtaining the magnitude of the difference between the length of the first integrated section and the average black pixel section length as the first integrated black pixel section length, and a process of obtaining the other side of the candidate section. When the second candidate black pixel section that is the nearest black pixel section and this candidate section are defined as one continuous black pixel section, the black pixel section is defined as a second integrated section, and the length of the second integrated section is defined as A process for determining the length of the difference between the length of the second integrated section and the average black pixel section length as a second integrated black pixel section length; a first integrated section length error and a second integrated section length error; And selecting the integrated section with the smaller error as the selected integrated section, and the length of the white pixel section between the candidate section of the selected integrated section and the nearest black pixel section of the candidate section And the length of the white pixel section If the length of the white pixel section is shorter than the fifteenth reference value, the line segmentation is corrected by setting the selected integrated section as one continuous black pixel section. And a process.

As described above, in the first line segmentation correcting method of the present invention, one line is mistakenly replaced with two lines in the same manner as in the case of the above (integrated correction 1) in the line segmenting method of the present invention. Correction of line cutout when cut out can be made. The 14th reference value may be set to any suitable value, similarly to the above-mentioned 6th reference value. Also, the fifteenth reference value may be set to any suitable value, similarly to the seventh reference value.

(Second Line Extraction Correction Method) In the above (division correction 1) and (division correction 2) in the line extraction method of the present invention, the division line extraction was corrected. The method is not limited to the modification of the extraction of the divided line, but can be applied to the modification of the extraction of the line which is not divided.

Therefore, according to the second line segmentation correcting method of the present invention, scanning is performed along a first direction substantially along the line direction of the document image to detect the cumulative number of black pixels for each scan. ,
A projection distribution is created by obtaining a distribution of the cumulative number of black pixels along a second direction orthogonal to the first direction. In the projection distribution, a projected portion of the cumulative number of black pixels equal to or larger than the sixteenth reference value is located in the second direction. After performing a line segmentation with a scanning region in which each continuous black pixel section is obtained as one row region,
In correcting the line segmentation, the process of obtaining the average value of the length of the black pixel section (black pixel section length) of the segmented row region, and comparing the average value with the length of the black pixel section to determine the black pixel section A process of selecting a black pixel section in which the ratio of the black pixel section length to the average value is larger than the seventeenth reference value as a division candidate section; Among the positions where the pixels are detected, the process of determining the position closest to the first side along the first direction as the first position, and the process of setting the first position closest to the other side along the first direction among the first positions. The process of determining the position as the first position and the position where the black pixel is detected on the scanning line that scans the vicinity of the center of the division candidate section in the scanning line.
A process of obtaining a position closest to the other side along the second direction as the second position, a process of obtaining a second position closest to the side of the second position along the second direction as the second position, By comparing the nearest first position and the nearest second position,
When the nearest first position is on the other side of the nearest second position and when the nearest first position and the nearest second position are located on the same straight line along the second direction. , Dividing the division candidate section into two in the center of the division candidate section to generate two divided black pixel sections, thereby correcting the line segmentation, the nearest first position and the nearest second position. When the nearest first position is located on one side of the nearest second position, the second first position passes through an intermediate point between the nearest first position and the nearest second position. By dividing the first division candidate section by a division line along the direction, the first divisional area on one side and the other divisional area on the other side are generated, and the second position closest to the one divisional area is obtained. It is divided by a scanning line to generate two divided black pixel sections, and the other divided area is placed at the nearest first place. Characterized in that it comprises a process for correcting the line cut by generating two divided black pixel period is divided by the scan line is obtained.

As described above, in the second line segmenting and correcting method of the present invention, two individual lines are erroneously changed to one line in the same manner as in the above-described (division modification 1) in the line segmenting method of the present invention. Correction of line cutout when cut out as a line can be performed. Further, as the sixteenth reference value, similarly to the above-described eleventh reference value, any suitable value may be set.

(Line clipping device) According to the line clipping device of the present invention, when the line direction of the document image is unknown or when the line direction of the document image is the first direction, the line is orthogonal to the first direction. A first direction divided region generating unit that generates a plurality of first direction divided regions arranged along the first direction by dividing the document image by dividing lines along the second direction; When an area is generated, each of the first direction divided areas is scanned in the first direction to detect the cumulative number of black pixels for each scan, and is then scanned along the second direction of the cumulative number of black pixels. A first-direction projection distribution creating unit for creating a first-direction projection distribution by calculating a first-direction projection distribution, and, when the first-direction projection distribution is created, a first reference value or more in each of the first-direction projection distributions A section in which the cumulative number of black pixels is continuous along the second direction is In the case where the scanning direction in which the first black pixel section is obtained is the first division row, a first division line extraction section that performs division line extraction as the first black pixel section, and the line direction of the document image is unknown And when the row direction of the document image is the second direction, by dividing the document image by a dividing line along the first direction, a plurality of lines arranged along a second direction orthogonal to the first direction. A second direction divided area generating unit for generating a second direction divided area, and, when the second direction divided area is generated, each of the second direction divided areas is scanned along the second direction, and each scan is performed. The second direction projection distribution is created by detecting the cumulative number of black pixels in the first direction and calculating the distribution of the cumulative number of black pixels along the first direction.
A direction projection distribution creating unit, and when a second direction projection distribution is created, a section in each of the second direction projection distributions in which the cumulative number of black pixels equal to or greater than a second reference value continues along the first direction. As a second black pixel section, a second divided line cutout unit that performs a divided line cutout using a scan area in which the second black pixel section is obtained as a second divided line, and a case where the line direction of the document image is unknown, A line direction determination unit that determines a line direction of the document image based on the first direction projection distribution and the second direction projection distribution, and a line direction determination unit that is adjacent to each other in the first direction when the line direction of the document image is the first direction The divided lines included in the first direction divided regions are connected to each other, and at least a part of the first black pixel section obtained by scanning the divided lines overlaps with each other to perform line segmentation. First split row connection to be performed When the line direction of the document image is the second direction, the divided lines included in the second direction divided areas adjacent to each other in the second direction are the second divided lines obtained by scanning the divided lines. And a second divided row connection unit that connects divided rows in which at least a part of the black pixel sections overlap each other to perform line segmentation.

As described above, according to the line segmenting apparatus of the present invention, even if the line direction of the document image is unknown, the line segmenting apparatus of the present invention is separately provided separately from the line segmenting method. Line cutout can be performed without performing the process of determining the line direction.

(Row Direction Determining Unit 1) In implementing the line extracting apparatus of the present invention, preferably, the row direction determining unit sets each of the first black pixel sections as a first black pixel section length. A first variation calculation unit that calculates the variation of the first black pixel section length, and sets the length of each of the second black pixel sections as the second black pixel section length, and calculates the variation of the second black pixel section length. The calculated second variation calculation unit compares the variation of the first black pixel section length with the variation of the second black pixel section length to determine the variation of the first black pixel section length and the variation of the second black pixel section length. When the ratio with the variation is equal to or more than the third reference value, a variation comparison / determination unit that determines the scanning direction of the projection distribution from which the black pixel section length with the smaller variation is obtained as the row direction may be provided.

As described above, if the row direction determining section (row direction determining section 1) is provided, the row direction determining method of the present invention described above (row direction determination based on variation in black pixel section length).
As described in the section, the determination of the row direction can be reliably performed.

(Line Direction Judgment Unit 2) Further, in implementing the line cutout apparatus of the present invention, preferably, the line direction judgment unit performs the first line-by-line judgment by focusing on the line interval and the character interval.
When the ratio of the variation of the black pixel section length to the variation of the second black pixel section length is less than the third reference value, the first white pixel between adjacent first black pixel sections in the first direction projection distribution. The row direction may be determined using the length of the section and the length of the second white pixel section between adjacent second black pixel sections in the second direction projection distribution.

By the way, even if the first and second black pixel section length variations are not compared, the line direction can be determined by paying attention to the line interval and the character interval.

Therefore, in implementing the line segmenting apparatus of the present invention, it is preferable that the line direction judging section sets the first direction between adjacent first black pixel sections in the first direction projection distribution.
It is preferable to determine the row direction using the length of the white pixel section and the length of the second white pixel section between adjacent second black pixel sections in the second direction projection distribution.

Further, in determining the row direction by using the lengths of the first and second white pixel sections, the row direction determination section determines whether the variation in the first black pixel section length and the second black pixel section length are different. If the ratio with the variation is less than the third reference value, and if there is one first white pixel section between adjacent first black pixel sections in the first direction projection distribution, the first white pixel When the length of the section is the first average white pixel section length, and when there are a plurality of the first white pixel sections, the arithmetic mean of the respective lengths of the first white pixel section is calculated as the first average white pixel section. When the first average white pixel section length calculation unit that calculates the pixel section length and the second white pixel section between adjacent second black pixel sections in the second direction projection distribution are one place, The length of the two white pixel sections is defined as a second average white pixel section length, and
When there are a plurality of white pixel sections, a second average white pixel section length calculation unit that calculates an arithmetic mean of the lengths of the second white pixel sections as a second average white pixel section length; An average value comparison / determination unit that compares the average white pixel section length with the second average white pixel section length and determines the scanning direction of the obtained projected distribution of the white pixel section having a larger average value as the row direction. It is good to get it.

As described above, if the row direction determining section (row direction determining section 2) is provided, the description has been given in the section of (row direction determination based on average white pixel section length) in the above-described row direction determining method of the present invention. As described above, it is possible to reliably determine the row direction.

(Line Direction Determining Unit 3) In implementing the line segmenting device of the present invention, preferably, the line direction determining unit obtains the total number of the first direction divided regions as the first divided region number. A division region number counting section, a first black pixel section number counting section for calculating the total number of first black pixel sections over all the first direction division regions as a first black pixel section number, and a second direction division region And a second divided area number counting unit that calculates the total number of the second divided pixel areas as the number of the second divided area, and a second divided pixel number that calculates the sum of the number of the second black pixel sections over all the second direction divided areas as the second black pixel area number. 2 black pixel section number counting section, the first black pixel section number is equal to the first divided area number, and
If the number of black pixel sections does not match the number of second divided areas, the row direction is determined to be the first direction, or the number of second black pixel sections and the number of second divided areas match, and If the number of 1 black pixel sections does not match the number of the first divided areas, the row direction is changed to the second
And a section number division number comparison / determination unit for determining the direction.

As described above, if the row direction determining section (row direction determining section 3) is provided, the description has been given in the section of (row direction determination for only one row) in the above-described row direction determining method of the present invention. As described above, when the number of lines included in the document image is one, the line direction can be easily determined.

(Division Row Connection) In implementing the line segmentation device of the present invention, preferably, the first division row connection is a part of the first black pixel section of the first direction division row included in the division region. Are respectively overlapped with a part of the first black pixel section of two first divided rows included in another divided area adjacent to the divided area, and the length of the overlapping section is the fourth reference. Connect split lines longer than the value,
The second divided row connection unit is configured such that a part of the second black pixel section of the second direction divided row included in the divided area includes two second divided rows included in another divided area adjacent to the divided area. When each overlaps with a part of the second black pixel section,
It is preferable to connect divided rows in which the length of the overlapping section is longer than the fifth reference value.

As described above, if the first divided row connection section and the second divided row connection section are provided, as described in the section (connection of divided rows) in the above-described row direction determination method of the present invention, Line segmentation can be performed by appropriately connecting the divided lines.

(First Integrated Correction Unit) In implementing the line extracting apparatus of the present invention, preferably, a first integrated correction unit for correcting line extraction is provided. A first average for calculating an arithmetic average of the length of the first black pixel section (hereinafter, also referred to as the first black pixel section length) of the first divided row cut out by the row cutout section as the first average black pixel section length. A black pixel section length calculation unit compares the first average black pixel section length with the first black pixel section length, and determines the first black pixel section of the first black pixel section with respect to the first average black pixel section length. A first candidate section selection unit that selects a first black pixel section whose length ratio is smaller than a sixth reference value as a first candidate section, and a first black pixel section closest to one side of the first candidate section. A certain first candidate black pixel section and this first candidate section are defined as one continuous black pixel section. And then the black pixel period when the first integration period, a first integration interval length detector for determining the length of the first integration interval, the length of the first integration period the first
A first integrated section length error calculating unit that obtains the magnitude of the difference from the average black pixel section length as a first integrated section length error, and a first black pixel section that is the nearest first black pixel section on the other side of the first candidate section. When the two candidate black pixel sections and the first candidate section are defined as one continuous black pixel section, the black pixel section is defined as a second integrated section, and the second integrated section length for calculating the length of the second integrated section A detecting unit, a second integrated section length error calculating unit that obtains a magnitude of a difference between the length of the second integrated section and the first average black pixel section length as a second integrated section length error, and a first integrated section length error And the second
A first selected integrated section selecting unit that compares the integrated section length error with the integrated section having a smaller error as the first selected integrated section; and a first candidate section of the first selected integrated section. A first white pixel section length detector for calculating a length of a first white pixel section between the first candidate pixel section and a first black pixel section immediately adjacent to the first candidate section; and a length of the first white pixel section and a seventh reference value. When the length of the first white pixel section is shorter than the seventh reference value, the first selection integrated section is set as one continuous first black pixel section, so that line segmentation is performed. Preferably, a first integration unit for performing correction is provided.

As described above, if the first integrated correction section is provided, it is possible to correct the divided line cutout as described in the section (integrated correction 1) in the above-described row direction determination method of the present invention. .

(Second Integrated Correction Unit) In implementing the line extracting apparatus of the present invention, preferably, a second integrated correction unit for correcting line extraction is provided. The second average for calculating the arithmetic average of the length of the second black pixel section (hereinafter also referred to as the second black pixel section length) of the second divided row cut out by the row cutout section as the second average black pixel section length. The black pixel section length calculation unit compares the second average black pixel section length with the second black pixel section length, and, among the second black pixel sections, the second black pixel section length with respect to the second average black pixel section length A second black pixel section that selects a second black pixel section whose ratio is smaller than the eighth reference value as a second candidate section, and a second black pixel section closest to one side of the second candidate section. The third candidate black pixel section and the second candidate section are defined as one continuous black pixel section. In this case, the black pixel section is referred to as a third integrated section, and a third integrated section length extraction unit for calculating the length of the third integrated section; and a third integrated section length and a second average black pixel section length. A third integrated section length error calculating unit that obtains the magnitude of the difference as a third integrated section length error, a fourth candidate black pixel section that is the second black pixel section closest to the other side of the second candidate section, and When the second candidate section is defined as one continuous black pixel section, the black pixel section is referred to as a fourth integrated section, and a fourth integrated section length extraction unit for calculating the length of the fourth integrated section; A fourth integrated section length error calculating section for obtaining a magnitude of a difference between the length of the integrated section and the second average black pixel section length as a fourth integrated section length error, a third integrated section length error and a fourth integrated section length error And selecting the integrated section having the smaller error as the second selected integrated section. And integrating section selecting unit, the second between the one of the second selection integration section, a second candidate section and the nearest second black pixel interval of the second candidate section
A second white pixel section length extraction unit for determining the length of the white pixel section;
By comparing the length of the second white pixel section with the ninth reference value,
When the length of the second white pixel section is shorter than the tenth reference value, the second selection integrated section is set as one continuous second black pixel section, thereby correcting the line segmentation.
It is good to have an integration unit.

As described above, if the second integrated correction unit is provided, it is possible to correct the divided line cutout as described in the column of (integrated correction 2) in the above-described row direction determination method of the present invention. .

(First Division Correction Unit) In implementing the line segmentation apparatus of the present invention, preferably, a first division modification unit for correcting line segmentation is provided, and the first division modification unit is provided with a first division modification unit. A first average for calculating an arithmetic average of a length of a first black pixel section (hereinafter, also referred to as a first black pixel section length) of a first divided row cut out by the row cutout unit as a first average black pixel section length. The black pixel section length calculation unit compares the first average black pixel section length with the first black pixel section length, and, among the first black pixel sections, the first black pixel section length with respect to the first average black pixel section length A first division candidate section selection unit that selects a first black pixel section having a ratio of greater than the eleventh reference value as a first division candidate section, and a scanning line that scans the vicinity of the center of the first division candidate section among the scanning lines Of the positions where black pixels are detected in
A first position extraction unit that determines a position closest to one side along the direction as a first position, and a first position extraction unit that determines a first position closest to the other side along the first direction as a first position among the first positions. A first position extracting unit that determines a position of the black line on the scanning line that scans the vicinity of the center of the first division candidate section among the scanning lines in the second direction and a position closest to the other side in the second direction;
A second position extracting unit to be obtained as a position;
By comparing the nearest second position with the nearest second position extracting unit that determines the nearest second position along the second direction as the nearest second position, it compares the nearest second position with the nearest second position. When the first position is on the other side of the nearest second position and when the nearest first position is
When the position and the nearest second position are located on the same straight line along the second direction, the first division candidate section is divided into two at the center of the first division candidate section to be divided into two. By generating a black pixel section, the first division unit that corrects line segmentation is compared with the nearest first position and the nearest second position, and the nearest first position is determined as the nearest second position. When the first division candidate section is located on one side of the position, the first division candidate section is divided by a division line passing through the middle between the nearest first position and the nearest second position and extending in the second direction. A first divided region on the side and the other divided region on the other side are generated, and the one divided region is divided by a scanning line from which the nearest second position is obtained to generate two divided black pixel sections, and The divided area is divided by the scanning line from which the nearest first position is obtained to generate two divided black pixel sections. It may become comprises a second division unit to correct the line cut by the.

As described above, if the first division correction unit is provided, it is possible to correct the division line cutout as described in the (division correction 1) section in the above-described row direction determination method of the present invention. .

(Second Division Correction Unit) In implementing the line segmentation apparatus of the present invention, preferably, a second division modification unit for correcting line segmentation is provided, and the second division modification unit is provided with a second division modification unit. The second average for calculating the arithmetic average of the length of the second black pixel section (hereinafter also referred to as the second black pixel section length) of the second divided row cut out by the row cutout section as the second average black pixel section length. The black pixel section length calculation unit compares the second average black pixel section length with the second black pixel section length, and, among the second black pixel sections, the second black pixel section length with respect to the second average black pixel section length And a second division candidate section selecting unit that selects a second black pixel section having a ratio greater than the twelfth reference value as a second division candidate section, and a scanning line that scans the vicinity of the center of the second division candidate section among the scanning lines. Out of the positions where black pixels are detected in
A third position extracting unit that determines a position closest to one side along the direction as the third position, and a third position extracting unit that determines a third position closest to the other side along the second direction among the third positions as the closest third position. A third position extracting unit that determines a position closest to the other side in the first direction among positions where black pixels are detected on a scanning line that scans the vicinity of the center of the second division candidate section among the scanning lines as a fourth position;
A fourth position extracting unit to be determined as a position;
The nearest fourth position extracting unit that determines the fourth position closest to the first position along the first direction as the nearest fourth position is compared with the nearest third position and the nearest fourth position, and the nearest fourth position is extracted. When the third position is on the other side of the nearest fourth position, the second division candidate section is divided into two at the center of the second division candidate section to generate two divided black pixel sections. Thus, the third division for correcting the line segmentation, the nearest third position and the nearest fourth
The third position is closer to the first position than the fourth position, and the third position and the fourth position are the same in the first direction. When the second candidate section is located on a straight line, the second candidate section is divided by a dividing line passing through the middle between the nearest third position and the nearest fourth position and extending along the first direction. A first divided region and another divided region on the other side are generated, and one divided region is divided by a scanning line at which a nearest fourth position is obtained to generate two divided black pixel sections, and 3rd nearest
It is preferable to include a fourth division unit that corrects line segmentation by generating two divided black pixel sections by dividing the scanning line whose position has been obtained.

As described above, if the second division correction unit is provided, the division line extraction can be corrected as described in the section (division correction 2) in the above-described row direction determination method of the present invention. .

(Line Direction Confirmation Switching Unit and Switching Output Unit) In implementing the line segmentation device of the present invention, it is preferable that the information of the document image externally input to the line segmentation device include the line direction of the document image. It is checked whether or not the information indicating that the line direction is the first direction is sent. If the information indicating that the line direction is the first direction is added, the information of the document image is transmitted to the first divided area generating unit. If the information indicating that the line direction is the second direction is added, the information of the document image is sent to the second divided area generating unit, or the information indicating the line direction is not added. In a case where the line direction confirmation switching unit that sends information of the document image to the first and second divided region generation units and the line direction determination unit determines that the line direction is the first direction, Outputs information obtained from the first segmented line segmentation unit , The row direction if it is determined that the second direction, may become comprises a switching output unit for outputting information obtained from the second division line cutout portion.

As described above, if the line direction confirmation switching unit is provided, it is possible to select the transmission destination of the information of the document image according to the information added to the document image. In addition, when the switching output unit is provided, when the row direction is determined, it is possible to obtain only information necessary for cutting out the row in the determined row direction.

Note that the line direction confirmation switching unit and the switching output unit may be external to the line cutout device.

(First Line Cutout Correcting Apparatus) According to the first line cutout correcting apparatus of the present invention, the document image is scanned in the first direction substantially along the line direction of the document image. The projection distribution is created by detecting the cumulative number of black pixels for each scan, and calculating the distribution of the cumulative number of black pixel sections along the second direction. A line cutout correction device for correcting a line cutout obtained by performing line cutout using a scanning region in which each black pixel section in which a number of projected portions are continuous in the second direction as one line region, and An average black pixel section length calculation unit for calculating an average black pixel section length of a black pixel section length (hereinafter, also referred to as a black pixel section length) of the selected row area; And compare the average black pixel A candidate section selection unit that selects a black pixel section in which the ratio of the black pixel section length to the interim length is smaller than the fourteenth reference value as a candidate section, and a first candidate that is the nearest black pixel section on one side of the candidate section When the black pixel section and the candidate section are set as one continuous black pixel section, the black pixel section is defined as a first integrated section, and a first integrated section length extraction unit that determines the length of the first integrated section; A first integrated section length error calculating section for obtaining the magnitude of the difference between the length of the first integrated section and the average black pixel section length as a first integrated section length error, and a nearest black pixel on the other side of the candidate section When the second candidate black pixel section, which is a section, and this candidate section are defined as one continuous black pixel section, the black pixel section is defined as a second integrated section, and the second integration for obtaining the length of the second integrated section is performed. The section length extraction unit, the length of the second integrated section and the average black A second integrated section length error calculating unit that obtains the magnitude of the difference from the elementary section length as a second integrated section length error is compared with the first integrated section length error and the second integrated section length error. A selection integration section selection unit that selects one of the integration sections as a selection integration section, and a white pixel that determines the length of a white pixel section between a candidate section of the selection integration section and the nearest black pixel section of the candidate section The section length extraction unit compares the length of the white pixel section with the fifteenth reference value. If the length of the white pixel section is shorter than the fifteenth reference value, the selected integrated section is converted into one continuous black It is characterized in that the pixel section is provided with an integration unit for correcting line segmentation.

As described above, according to the first line segmentation correcting apparatus of the present invention, when one line is mistakenly segmented as two lines, the first line segmentation correcting method of the present invention described above. In the same manner as described above, the line cutout can be corrected.

(Second Line Extraction Correction Device) According to the second line extraction correction device of the present invention, a document scans a layer along a first direction substantially along a line direction of a document image. Then, a projection distribution is created by detecting the cumulative number of black pixels for each scan, and calculating the distribution of the cumulative number of black pixels along a second direction orthogonal to the first direction.
A line segmentation that corrects the segmentation of a line obtained by performing line segmentation using a scanning area in which each black pixel section in which a projected portion having a cumulative number of black pixels equal to or more than the reference value is continuous in the second direction as one row area. An average black pixel section length calculation unit for obtaining an average black pixel section length of a black pixel section length (hereinafter, also referred to as a black pixel section length) of a cut-out row region; The length of the black pixel section is compared with the length of the black pixel section, and a black pixel section in which the ratio of the black pixel section length to the average black pixel section length is greater than the seventeenth reference value is selected as a division candidate section. A section selector,
A first position extracting unit that determines, as a first position, a position closest to the one side along the first direction among positions where black pixels are detected on a scanning line that scans the vicinity of the center of the division candidate section among the scanning lines; Among the first positions, a nearest first position extracting unit that determines the nearest first position along the first direction as the nearest first position, and a scan that scans the vicinity of the center of the division candidate section in the scan line. Among the positions where black pixels are detected on the line, a second position extracting unit that determines the position closest to the other side in the second direction as the second position, and one of the second positions along the second direction. By comparing the nearest second position with the nearest second position extracting unit that determines the nearest second position as the nearest second position, the nearest first position is compared with the nearest second position. When the second position is located on the other side of the second position and the nearest first position and the nearest second position are in the second direction. Is divided into two at the center of the division candidate section to generate two division black pixel sections, thereby correcting the line segmentation. The division unit compares the nearest first position and the nearest second position. If the nearest first position is one side of the nearest second position, the nearest first position and the nearest first position are compared with each other. By dividing the first division candidate section by a division line passing through the middle with the nearest second position and along the second direction, one divided area on one side and the other divided area on the other side are generated. The divided area is divided by the scanning line at which the nearest second position is obtained to generate two divided black pixel sections, and the other divided area is divided by the scanning line at which the nearest first position is obtained. Second division unit that corrects line segmentation by generating two divided black pixel sections Characterized by comprising comprises a.

As described above, according to the second line cutout correcting device of the present invention, when two lines are cut out as one line by mistake, the above-mentioned second line cutout correcting device of the present invention is used. In the same manner as in the method, the line segmentation can be corrected.

[0089]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a first embodiment of the present invention;
An embodiment of a line segmentation device and first and second line segmentation correction devices will be described together. The present invention is not limited to the illustrated example.

(Constitution) First, referring to FIG.
The line cutting device according to the present invention will be described. FIG. 1 is a functional block diagram for explaining the configuration of the line segmenting apparatus according to this embodiment. A detailed description of the operation of the line segmenting device 10 will be described later in the section (Operation).

The line segmenting apparatus 10 of this embodiment includes a line direction confirmation switching unit 12, a first direction divided area generation unit 14a,
Second direction divided area generation unit 14b, first direction projection distribution creation unit 16a, second direction projection distribution creation unit 16b, first divided line cutout unit 18a, second divided line cutout unit 18b, line direction determination unit 20, switching An output unit 22, a first correction unit 24a, a second correction unit 24b, a first divided row connection unit 26a, a second divided row connection unit 26b, and a control unit 28 are provided.

The row direction confirmation switching section 12 is constituted by a row direction confirmation section 12a and a switching section 12b. The line direction checking unit 12a checks whether information indicating the line direction of the document image is added to the information of the document image input from the outside to the line cutout device.

Here, arbitrary directions orthogonal to each other can be set as the first direction and the second direction. For example, if the first direction is a horizontal direction (horizontal direction),
The second direction is a vertical direction (vertical direction). Further, the first direction may be the vertical direction and the second direction may be the horizontal direction.

When the line direction confirmation unit 12a confirms that the information indicating that the line direction of the document image is the first direction is added to the information of the document image, the switching is performed. The unit 12b transmits the information of the document image together with the information (confirmation result) to the first direction divided area generation unit 14
Output to a.

When the line direction confirmation unit 12a confirms that information indicating that the line direction of the document image is the second direction is added to the information of the document image, The switching unit 12b outputs the information of the document image together with the information (confirmation result) to the second direction divided area generation unit 14
Output to b.

When the line direction confirmation unit 12a confirms that the information indicating the line direction of the document image is not added to the information of the document image, the switching unit 12b
Outputs the information of the document image to both the first and second direction divided area generation units 14a and 14b.

Further, the first direction divided area generating section 14a
When the line direction of the document image is unknown and when the line direction of the document image is the first direction, the document image is divided by the dividing line along the second direction, so that the document images are arranged in the first direction. A plurality of first direction divided areas are generated. Then, the first direction divided area generation unit 14a transmits the generated information of the first direction divided area to the first direction projection distribution generation unit 16a.
a.

The first direction projection distribution creating unit 16a
When the first direction divided area is generated, each of the first direction divided areas is scanned along the first direction to detect the cumulative number of black pixels for each scan, and the first cumulative number of black pixels is calculated. A first direction projection distribution is created by obtaining a distribution along a second direction orthogonal to the direction. That is, when the first direction divided area is generated, the first direction projection distribution creating unit 16a sets the first direction as the main scanning direction and the second direction as the sub scanning direction for each of the first direction divided areas. The first direction projection distribution is created by scanning the first direction divided area and detecting the cumulative number of black pixels for each scan. Then, the first direction projection distribution creating unit 16a sends the created information of the first direction projection distribution to the first divided row cutout unit 18a.

The first divided line segmenting section 18a
When the directional projection distribution is created, the cumulative number of black pixels equal to or more than the first reference value in each of the first directional projection distributions is equal to the second reference projection distribution.
A section that is continuous along the direction is defined as a first black pixel section, and a scanning region in which the first black pixel section is obtained is defined as a first division row, and a divided line is cut out. When the row direction is the first direction, the first divided row cutout unit 18a adds the information of the cut divided row to the first correction unit 24a.
Send to Also, the first divided line segmentation unit 18a
If the row direction is unknown, information on the cut-out divided rows is sent to the row direction determination unit 20.

Further, the second direction divided area generating section 14b
When the line direction of the document image is unknown or when the line direction of the document image is the second direction, the document image is divided by a dividing line along the first direction, whereby the first
A plurality of second direction divided regions arranged along a second direction orthogonal to the direction are generated. Then, the second direction divided area generation unit 14b sends the generated information of the second direction divided area to the second direction projection distribution creation unit 16b.

The second direction projection distribution creating unit 16b
When the second direction divided area is generated, each of the second direction divided areas is scanned along the second direction to detect the cumulative number of black pixels for each scan, and to determine the number of the cumulative black pixels. A second direction projection distribution is created by obtaining a distribution along one direction. That is, when the second direction divided area is generated, the second direction projection distribution creating unit 16b sets the second direction as the main scanning direction and the first direction as the sub scanning direction for each of the second direction divided areas. The second direction divided area is scanned to detect the cumulative number of black pixels for each scan, thereby creating a second direction projection distribution. Then, the second direction projection distribution creating unit 16b stores the information of the created second direction projection distribution in the second direction.
The data is sent to the divided line cutout unit 18b.

Further, the second divided row cutout section 18b
When the directional projection distribution is created, the cumulative number of black pixels equal to or greater than the second reference value in each of the second directional projection distributions is the first.
A section that is continuous along the direction is defined as a second black pixel section, and a scanning area in which the second black pixel section is obtained is defined as a second divided row, and divided row cutout is performed. When the row direction is the second direction, the second divided row cutout unit 18b adds the information of the cut second divided row to the second correction unit 2b.
4b. Also, the second divided line cutout unit 18b
Sends the information of the cut-out second divided row to the row direction determination unit 20 when the row direction is unknown.

Further, when the line direction of the document image is unknown, the line direction determination unit 20 determines the line direction of the document image based on the first direction projection distribution and the second direction projection distribution.
Then, the row direction determination unit 20 determines the first row
The information on the divided row and the second divided row is sent to the switching output unit 22. Alternatively, the row direction determination unit 20 may transmit only the determination result of the row direction and the information on the determined divided row in the row direction to the change-over output unit 22.

When the determined row direction is the first direction, the switching output unit 22 outputs the information of the first divided row to the first direction.
It is sent to the correction unit 24a. Also, the switching output unit 22
Is the second direction if the determined row direction is the second direction.
The information of the divided row is sent to the second correction unit 24b.

Further, the first correction unit 24a corrects the division line extraction when the first division line is erroneously extracted. Then, the corrected information of the first divided row is sent to the first divided row connecting unit 26a.

When the row direction of the document image is the first direction, the first divided row connecting section 26a is used for dividing rows included in the first direction divided areas adjacent to each other in the first direction. The line segmentation is performed by connecting the divided lines in which at least a part of the first black pixel section obtained by scanning the divided line overlaps.

By the way, when connecting each divided row, there is a case where one divided row can be connected to two divided rows by only one side. That is, the black pixel section of one divided row may partially overlap the black pixel section of two divided rows on one side only.

Therefore, in the first divided row connecting section 26a of this embodiment, a part of the first black pixel section of the first direction divided row included in the divided area is replaced with another one of the other adjacent ones of the divided area. When each of the two first divided rows included in the divided area overlaps with a part of the first black pixel section, the divided rows whose overlapping sections are longer than the fourth reference value are connected.

Then, the first divided row connection unit 26a sends the information of the cut-out row to the control unit 28.

The fourth reference value may be set to any suitable value. For example, it is preferable that the fourth reference value is set so that the amount of deviation between the divided rows connected to each other is minimized. Further, for example, when the black pixel section length and the white pixel section length are almost the same, the fourth reference value may be set to a value of about “0.5”. If the fourth reference value is set to such a value, if one of the two divided rows of the adjacent divided area overlapping on one side of one divided row satisfies the reference value, the other is the reference value. Will not be satisfied.

Further, the second correction section 24b corrects the division line extraction when the second division line is erroneously extracted. Then, the corrected second divided row is sent to the second divided row connecting unit 26b.

When the row direction of the document image is the second direction, the second divided row connecting section 26b is used to connect the divided rows included in the second direction divided areas adjacent to each other in the second direction. The line segmentation is performed by connecting the divided lines in which at least a part of the second black pixel section obtained by scanning the divided line overlaps.

In the process of connecting the second divided rows, the second divided row connecting section 26b converts a part of the second black pixel section of the second direction divided row included in the divided area into the divided area. In the case where each overlaps with a part of the second black pixel section of the two second divided rows included in another adjacent divided area, the length of the overlapping section is the fifth.
Connect divided lines longer than the reference value.

Then, the second divided row connecting section 26b sends the information of the cut-out rows to the control section 28.

The fifth reference value may be set to any suitable value, similarly to the fourth reference value. For example, the fourth reference value and the fifth reference value may be equal to each other, or may be different from each other.

The control unit 28 controls the operation of each unit of the line segmenting device 10. Then, the control unit 28 outputs the line segmentation result to the outside of the line segmentation device 10. FIG.
2, the connection between the control unit 28 and each unit of the line segmentation device 10 is partially omitted.

Therefore, according to this line segmenting device 10,
When the line direction of the document image is unknown, the line direction is determined using the first and second direction projection distributions obtained in the process of generating the divided area as a part of the line segmentation process. Therefore, according to the line cutout apparatus 10, even when the line direction of the document image is unknown, the line cutout can be performed without performing the process of determining the line direction separately from the line cutout in advance.

(Example of Configuration of Character Recognition Apparatus) Next, an example of the configuration of a character recognition apparatus using the line segmenting apparatus 10 of this embodiment will be described with reference to FIG. FIG. 2 is a functional block diagram for explaining a configuration example of the character recognition device.

This character recognition device includes a scanner 30, an image memory 31, a character area setting section 32, a line segmentation device 1
0, a character extracting section 33 and a character recognizing section 34 are provided.

The scanner 30 creates a document image (image data) from the document to be recognized. That is, the scanner 30 optically scans the paper surface of a document to generate image data of a document image including a quantized electric signal, for example, an electric signal representing black and white binary. In this image data, character pixels are represented by black pixels, and background pixels are represented by white pixels.

The image memory 31 temporarily stores the document image generated by the scanner 30.

The character area setting section 32 first reads out the document image stored in the image memory 31. Then, the character area setting section 32 sets a character area as an area for which the character direction is to be determined in the document image. When setting the character area, for example,
The image data may be displayed on a display, and the operator may specify a character area on the display. Also,
The inventor of the present application proposes an example of a method for extracting a character area in Document 2: "Japanese Patent Laid-Open No. 3-126181". According to the method disclosed in Document 2,
A circumscribed rectangle of the character pixel region is obtained, and a character region is cut out from the document image by using the circumscribed rectangle. In this embodiment, the character region may be extracted by using any other conventionally known and suitable method. Further, in the present invention, it is not always necessary to set a character area for line segmentation.

The line segmenting apparatus 10 has the configuration shown in FIG. 1, and operates as described later in the section of (Operation).

The character extracting section 33 extracts a character pattern from each line extracted by the line extracting device 10. In cutting out the character pattern, any conventionally known and suitable method can be used. In addition, the character extracting section 33 may include a built-in memory for temporarily storing the extracted character pattern.

The character recognizing unit 34 compares the cut character pattern with character dictionary data by a pattern matching method, and recognizes the character pattern based on the result of the comparison. Note that the dictionary data may be stored in a built-in memory of the character recognition unit 34, or a character recognition dictionary unit may be provided independently of the character recognition unit 34 and held there.

(Regarding the Row Direction Judgment Unit) Next, the configuration of the row direction judgment unit 20 in the line segmenting apparatus 10 of this embodiment will be described with reference to FIG. The row direction determining unit 20 includes a first determining unit 201, a second determining unit 202, and a third determining unit 203. Next, the configuration of each determination unit will be sequentially described.

(Regarding Configuration of First Determination Unit) Next, FIG.
The configuration of the first determination unit 201 will be described with reference to FIG. FIG. 4 is a functional block diagram for explaining the configuration of the first determination unit 201. The operation of the first determination unit 201 will be described later in detail in the section (Operation of the first determination unit).

The first determination unit 201 makes a line direction determination when a document image includes only one line of a character string.
The first determination unit 201 includes a first divided area number counting unit 36a, a second divided area number counting unit 36b, a first black pixel section number counting unit 38a, a second black pixel section number counting unit 38b, and a section number. A division number comparison / determination unit 40 is provided.

The first divided area number counting section 36a
Calculates the total number of the first direction divided areas as the first divided area number. Then, the first divided area number counting section 36a sends information on the counted number of first direction divided areas to the section number divided number comparison / determination section 40.

Further, the first black pixel section number counting section 38a
The total number of the first black pixel sections over all the first direction divided areas is determined as the first black pixel section number. And
The first black pixel section number counting section 38 a sends information on the counted first black pixel section number to the section number division number comparison determination section 40.

The second divided area number counting section 36b calculates the total number of divided areas in the second direction as the number of second divided areas. Then, the second divided region number counting unit 36b compares the counted information on the number of second divided regions with the section number divided number comparison determination unit 40b.
Send to

Further, the second black pixel section number counting section 38b
The sum of the number of second black pixel sections over all the second direction divided areas is determined as the number of second black pixel sections. And
The second black pixel section number counting section 38 b sends the information on the counted number of second black pixel sections to the section number division number comparison determination section 40.

The section number division number comparison / judgment unit 40 determines that the first black pixel section number matches the first division area number, and that the second black pixel section number does not match the second division area number. In the case of, the row direction is determined to be the first direction, and the number of the second black pixel sections is equal to the number of the second divided areas, and the number of the first black pixel sections and the number of the first divided areas are different. If they do not match, the row direction is determined to be the second direction.

As described above, the first determination unit 201 compares the number of divided areas with the total number of divided lines, and when the character string is only one line, the second determination unit 202
In addition, the third determination unit 203 can determine the row direction without having to compare variations in black pixel section lengths and average white pixel section lengths.

(Regarding the Configuration of the Second Judgment Unit) Next, FIG.
The configuration of the second determination unit 202 will be described with reference to FIG. FIG. 5 is a functional block diagram for explaining the configuration of the second determination unit 202. The operation of the second determination unit 202 will be described later in detail in the section (Operation of the second determination unit).

As one of the clues for the determination of the row direction, the second determination unit 202 pays attention to the variation in the row height and the row width. Make a decision. Then, the second determination unit 202 includes a first variation calculation unit 42a and a second
A variation calculation unit 42b and a variation comparison determination unit 44 are provided.

The first variation calculator 42a
Calculates the variation of the first black pixel section length, using each length of the first black pixel section as the first black pixel section length. Then, the first variation calculation unit 42a sends information on the calculated variation of the first black pixel section length to the variation comparison determination unit 44.

The second variation calculator 42b calculates the second
The length of each black pixel section is defined as a second black pixel section length, and the variation of the second black pixel section length is calculated. Then, the second variation calculation unit 42b compares the information on the variation of the calculated second black pixel section length with the variation comparison determination unit 44.
Send to

The variation comparison / determination section 44 compares the variation in the first black pixel section length with the variation in the second black pixel section length, and determines the variation in the first black pixel section length and the variation in the second black pixel section length. When the ratio with respect to the variation is equal to or more than the third reference value, the scanning direction of the projection distribution from which the black pixel section length with the smaller variation is obtained is determined as the row direction.

As described above, by comparing the variation in the first black pixel section length with the variation in the second black pixel section length, the determination in the row direction can be assured.

(Regarding Configuration of Third Judgment Unit) Next, FIG.
The configuration of the third determination unit 203 will be described with reference to FIG. FIG. 6 is a functional block diagram for explaining the configuration of the third determination unit 203. The operation of the third determination unit 203 will be described later in detail in the section (Operation of the third determination unit).

When the ratio of the variation in the first black pixel section length to the variation in the second black pixel section length is less than the third reference value in the second determination section 202, As one of the clues to determine the direction, the line direction is determined by averaging the intervals between black pixel sections in the projection distribution by focusing on the line interval and the character interval. The third determination unit 203 includes a first average white pixel section length calculation unit (first average value calculation unit) 46a and a second average white pixel section length calculation unit 46a.
An average white pixel section length calculation unit (second average value calculation unit) 46b and an average value comparison determination unit 48 are provided.

When the first average white pixel section length calculating section 46a has one first white pixel section between adjacent first black pixel sections in the first direction projection distribution, the first average white pixel section length calculation section 46a determines whether When the length of the first white pixel section is a first average white pixel section length, and when there are a plurality of first white pixel sections, the arithmetic average of the respective lengths of the first white pixel section is calculated. First
It is calculated as the average white pixel section length. Then, the first average white pixel section length calculation unit 46a sends the calculated first white pixel section length to the average value comparison determination unit 48.

The second average white pixel section length calculator 46b
In the case where there is one second white pixel section between adjacent second black pixel sections in the second direction projection distribution, the length of the second white pixel section is set to the second average white pixel section length. age,
In addition, when there are a plurality of second white pixel sections, the arithmetic mean of the lengths of the second white pixel sections is calculated as the second average white pixel section length. Then, the second average white pixel section length calculation unit 46b sends the calculated second white pixel section length to the average comparison determination unit 48.

The average comparison / determination section 48 compares the first average white pixel section length with the second average white pixel section length, and determines the scanning direction of the projection distribution having the larger average white pixel section length. Judge as line direction.

As described above, the line spacing is generally wider than the character spacing. Therefore, of the first average white pixel section length and the second average white pixel section length, the larger average white pixel section length corresponds to a blank between rows. Therefore, if the row direction is determined by comparing the first average white pixel section length with the second average white pixel section length, the ratio of the variation of the first black pixel section length to the variation of the second black pixel section length is determined. Even in a small case (for example, when the character interval is a fixed pitch), it is possible to ensure the determination of the line direction.

(Regarding the Configuration of the First Correction Unit) Next, FIG.
The configuration of the first correction unit 24a will be described with reference to FIG. FIG. 7 is a functional block diagram for explaining the configuration of the first correction unit. The first correction unit 24a includes the first integrated correction unit 9
0a and a first division correction unit 92a. The information input to the first correction unit 24a is input to the first integrated correction unit 90a and the first division correction unit 92a, respectively. Further, the first integrated correction unit 90a and the first division correction unit 92a
Are sent to the first divided row connection unit 26a.

Note that the first integrated correction unit 90a is also an embodiment of the first line segmentation correction device of the present invention. The first division correction unit 92a is also an embodiment of the second line segmentation correction device of the present invention.

(Regarding Configuration of First Integrated Correction Unit)
With reference to FIG. 8, the configuration of the first integrated correction unit 90a will be described. FIG. 8 is a functional block diagram for explaining the configuration of the first integrated correction unit 90a. First integrated correction unit 90a
Focuses on the height or width and the line interval of the first divided line, and finds a divided line that is erroneously cut out as two divided lines where it should be originally cut out as one divided line. Then, the first integrated correction unit 90a corrects the line extraction of the first divided line by cutting out the two divided lines as one divided line.

For this purpose, the first integrated correcting section 90a includes a first average black pixel section length calculating section 120a, a first candidate section selecting section 122a, a first integrated section length detecting section 124a, and a second integrated section length detecting section 124b. , First integrated section length error calculator 12
5a, a second integrated section length error calculating section 125b, a first selected integrated section selecting section 126a, a first white pixel section length detecting section 128
a and the first integration unit 130a.

The first average black pixel section length calculation section 120a calculates the length of the first black pixel section (first black pixel section length) of the first division row cut out by the first division row / line extraction section 18a. ) Is determined as the first average black pixel section length. Then, the first average black pixel section length calculation unit 120
a sends out the information of the first average black pixel section length to the first candidate section selection unit 122a.

The first candidate section selection unit 122a compares the first average black pixel section length with the first black pixel section length,
In the first black pixel section, the ratio of the first black pixel section length to the first average black pixel section length is smaller than the sixth reference value.
A black pixel section is selected as a first candidate section. Then, the first candidate section selection unit 122a sends information on the selected first candidate section to the first and second integrated section length detection units 124a and 124b, respectively. The sixth reference value may be set to any suitable value (for example, about 0.5) that is less than 1 and greater than 0, for example.

Further, the first integrated section length detecting section 124a
A first candidate black pixel section which is a first black pixel section nearest to one side of the first candidate section, and the black pixel section when the first candidate section is defined as one continuous black pixel section. 1
As an integrated section, the length of the first integrated section (first integrated section length) is obtained. Then, the first integrated section length detection unit 12
4a sends the obtained information of the first integrated section length to the first integrated section length error calculating section 125a.

The first integrated section length error calculator 125a
Calculates the magnitude of the difference between the length of the first integrated section and the first average black pixel section length as the first integrated section length error. Then, the first integrated section length error calculating section 125a uses the calculated first integrated section length error as the first selected integrated section selecting section 12a.
6a.

Further, the second integrated section length detecting section 124b
When the second candidate black pixel section, which is the nearest first black pixel section on the other side of the first candidate section, and this first candidate section are defined as one continuous black pixel section, 2
As an integrated section, the length of the second integrated section (second integrated section length) is obtained. Then, the second integrated section length detecting unit 12
4b sends the obtained information on the second integrated section length to the second integrated section length error calculating section 125b.

The second integrated section length error calculator 125b
Calculates the magnitude of the difference between the length of the second integrated section and the first average black pixel section length as the second integrated section length error. Then, the second integrated section length error calculating section 125b calculates the calculated second integrated section length error by the first selected integrated section selecting section 12b.
6a.

Also, the first selection integrated section selection section 126a
Compares the first integrated section length with the second integrated section length and selects the shorter integrated section as the first selected integrated section. Then, the first selected integrated section selection unit 126a sends information on the selected first selected integrated section to the first white pixel section length detection unit 128a.

Also, the first white pixel section length detecting section 128a
Calculates the length of the first white pixel section between the first candidate section and the first black pixel section immediately adjacent to the first candidate section in the first selection integration section. Then, the first white pixel section length detection unit 128a sends the obtained information on the length of the first white pixel section to the first integration unit 130a.

The first integration section 130a compares the length of the first white pixel section with the seventh reference value, and determines that the length of the first white pixel section is shorter than the seventh reference value. Then, the line selection is corrected by setting the first selected integrated section as one continuous first black pixel section. The seventh reference value may be set to any suitable value. For example, a value smaller than the average value of the length of the white pixel section (for example, about half the average value) may be set as the seventh reference value.

If the length of the first white pixel section is equal to or longer than the seventh reference value, it is highly likely that the first selected integrated section is not originally one line, so the line is cut out from the first candidate section. Is corrected to the line segmentation of the first selected integrated section.

(Regarding Configuration of First Division Correction Unit)
With reference to FIG. 9, the configuration of the first division correction unit 92a will be described. FIG. 9 is a functional block diagram for explaining the configuration of the first division correction unit 92a. First division correction unit 92a
Focuses on the height or width of the first divided line, and finds a divided line that is erroneously cut out as one divided line where it should be originally cut out as two divided lines. Then, the first division correction unit 92a adds the divided row to 2
By cutting out as one divided line, the line cutout of the first divided line is corrected.

For this purpose, the first division correction section 92a includes a first average black pixel section length calculation section 131a, a first division candidate section selection section 132a, a first position extraction section 134a, a second position extraction section 134b, First position extraction unit 136a, nearest second
A position extracting unit 136b, a first dividing unit 138a, and a second dividing unit 138b are provided.

The first average black pixel section length calculation section 131a calculates the length of the first black pixel section (first black pixel section length) of the first division row cut out by the first division row extraction section 18a. Is determined as a first average black pixel section length. Then, the first average black pixel section length calculation unit 131a calculates
The obtained information on the first average black pixel section length is sent to the first division candidate section selection unit 132a.

Further, the first division candidate section selection section 132a
Compares the first average black pixel section length with the first black pixel section length, and determines that the ratio of the first black pixel section length to the first average black pixel section length in the first black pixel section is equal to an eleventh standard. A first black pixel section larger than the value is selected as a first division candidate section. Then, the first division candidate section selection unit 132a sends information on the selected first division candidate section to the first and second position extraction units 134a and 134b. In addition, this first
One reference value is, for example, one or more arbitrary suitable values (for example, 1.
About 5 to 2).

In addition, the first position extracting unit 134a is the most one-side position along the first direction among the positions where black pixels are detected on the scanning line that scans the vicinity of the center of the first division candidate section in the scanning line. Is extracted as the first position. And
The first position extracting unit 134a sends information on the extracted first position to the nearest first position extracting unit 136a.

The nearest first position extracting section 136a extracts the first position closest to the other side in the first direction among the first positions as the nearest first position. Then, the nearest first position extracting unit 136a sends the extracted information on the nearest first position to the first and second dividing units 138a and 138b.

In addition, the second position extracting unit 134b moves the other position along the second direction along the second direction among the positions where black pixels are detected on the scanning line that scans the vicinity of the center of the first division candidate section in the scanning line. Is extracted as the second position. And
The second position extracting unit 134b sends the information on the extracted second position to the nearest second position extracting unit 136b.

The nearest second position extracting unit 136b extracts the second position closest to the first position along the second direction from the second positions as the nearest second position. Then, the nearest second position extracting unit 136b sends the extracted information on the nearest second position to the first and second dividing units 138a and 138b.

The first division 138a compares the nearest first position with the nearest second position, and determines whether the nearest first position is on the other side of the nearest second position or When the nearest first position and the nearest second position are located on the same straight line along the second direction, the first division candidate section is divided into two at the center of the first division candidate section. By generating two divided black pixel sections, the line segmentation is corrected.

The second division 138b compares the nearest first position with the nearest second position, and determines that the nearest first position is one side of the nearest second position. By dividing the first division candidate section by a division line that passes through the middle between the nearest first position and the nearest second position and extends in the second direction, the one division area on one side and the other division area on the other side are divided. The other divided area is generated, and one divided area is divided by the scanning line from which the nearest second position is obtained to generate two divided black pixel sections, and the other divided area is obtained as the nearest first position. The line segmentation is performed by dividing the scan line into two divided black pixel sections, thereby correcting the line segmentation.

(Regarding the Configuration of the Second Correction Unit) Next, FIG.
The configuration of the second correction unit 24b will be described with reference to FIG. FIG. 10 is a functional block diagram for explaining the configuration of the second correction unit. The second correction unit 24b includes a second integrated correction unit 90b and a second division correction unit 92b. The information input to the second correction unit 24b is transmitted to the second integrated correction unit 90b.
And the second division correction unit 92b. Further, a second integrated correction unit 90b and a second division correction unit 92b
Are sent to the second divided row connection unit 26b.

Note that the second integrated correction unit 90b is also an embodiment of the first line segmentation correction device of the present invention. The second division correction unit 92b is also an embodiment of the second line segmentation correction device of the present invention.

(Regarding Configuration of Second Integrated Correction Unit)
With reference to FIG. 11, the configuration of the second integrated correction unit 90b will be described. FIG. 11 is a functional block diagram for explaining the configuration of the second integrated correction unit 90b. Second integrated correction unit 9
0b focuses on the height or width and the line interval of the cut-out second divided line, and finds a divided line that is erroneously cut out as two divided lines where it should originally be cut out as one divided line. . Then, the second integrated correction unit 90b
By cutting out the two divided lines as one divided line, the line extraction of the second divided line is corrected.

For this purpose, the second integrated correcting section 90b includes a second average black pixel section length calculating section 120b, a second candidate section selecting section 122b, a third integrated section length detecting section 124c, and a fourth integrated section length detecting section 124d. , The third integrated section length error calculator 12
5c, fourth integrated section length error calculating section 125d, second selected integrated section selecting section 126b, second white pixel section length detecting section 128
b and a second integration unit 130b.

The second average black pixel section length calculation section 120b calculates the length of the second black pixel section (second black pixel section length) of the second division row cut out by the second division row extraction section 18b. Is determined as a second average black pixel section length. Then, the second average black pixel section length calculation unit 120b
Sends information on the second average black pixel section length to the second candidate section selection unit 122b.

The second candidate section selection unit 122b compares the second average black pixel section length with the second black pixel section length, and
In the second black pixel section, the ratio of the second black pixel section length to the second average black pixel section length is smaller than the eighth reference value.
The black pixel section is selected as the second candidate section. Then, the second candidate section selection unit 122b sends information on the selected second candidate section to the third and fourth integrated section length detection units 124c and 124d, respectively. The eighth reference value may be set to any suitable value (for example, about 0.5) smaller than 1 and larger than 0, for example, similarly to the above-mentioned sixth reference value. The eighth reference value and the sixth reference value may be equal to each other, or may be different values.

Further, the third integrated section length detecting section 124c
A second candidate black pixel section, which is the second black pixel section closest to one side of the second candidate section, and the black pixel section when the second candidate section is defined as one continuous black pixel section. 3
As an integrated section, the length of the third integrated section (third integrated section length) is obtained. Then, the third integrated section length detecting unit 12
4c sends the information on the obtained third integrated section length to the third integrated section length error calculating section 125c.

The third integrated section length error calculating section 125c
Calculates the difference between the length of the third integrated section and the second average black pixel section length as the third integrated section length error. Then, the third integrated section length error calculating section 125c sends out the calculated third integrated section length error to the second selected integrated section selecting section 126b.

Further, the fourth integrated section length detecting section 124d
When the fourth candidate black pixel section, which is the second black pixel section nearest to the other side of the second candidate section, and this second candidate section are defined as one continuous black pixel section, 4
As an integrated section, the length of the fourth integrated section (fourth integrated section length) is obtained. Then, the fourth integrated section length detector 12
4d sends the obtained information on the fourth integrated section length to the fourth integrated section length error calculating unit 125d.

The fourth integrated section length error calculating section 125d
Calculates the difference between the length of the fourth integrated section and the second average black pixel section length as the fourth integrated section length error. Then, the fourth integrated section length error calculating section 125d sends the calculated fourth integrated section length error to the second selected integrated section selecting section 126b.

The second selection integrated section selection section 126b
Compares the third integrated section length error with the fourth integrated section length error, and selects the integrated section having the smaller error as the second selected integrated section. Then, the second selection integrated section selection unit 126
b sends the information of the selected second selected integrated section to the second white pixel section length detecting unit 128b.

The second white pixel section length detecting section 128b
Calculates the length of the second white pixel section between the second candidate section and the second black pixel section immediately adjacent to the second candidate section in the second selection integration section. Then, the second white pixel section length detection unit 128b sends the obtained information on the length of the second white pixel section to the second integration unit 130b.

The second integration section 130b compares the length of the second white pixel section with the ninth reference value, and determines that the length of the second white pixel section is shorter than the ninth reference value. Next, the line selection is corrected by setting the second selected integrated section as one continuous second black pixel section. The ninth reference value may be set to any suitable value, similarly to the above-described seventh reference value. For example, the ninth reference value and the seventh reference value may be equal to each other, or may be different from each other.

If the length of the second white pixel section is equal to or longer than the ninth reference value, it is highly likely that the second selected integrated section is not originally one line, so the line is cut out from the second candidate section. Is corrected to the line segmentation of the second selected integrated section.

(Regarding the Configuration of the Second Division Correction Unit)
With reference to FIG. 12, the configuration of the second division correction unit 92b will be described. FIG. 12 is a functional block diagram for explaining the configuration of the second division correction unit 92b. Second integrated correction unit 9
2b focuses on the height or width of the cut-out second divided line, and finds a divided line that is erroneously cut out as one divided line where it should be cut out as two divided lines. Then, the second division correction unit 92b corrects the line division of the first division line by cutting out the division line as two division lines.

For this purpose, the second division correction section 92b includes a second average black pixel section length calculation section 131b, a second division candidate section selection section 132b, a third position extraction section 134c, a fourth position extraction section 134d, Third position extracting unit 136c, nearest fourth position
A position extracting unit 136d, a third dividing unit 138c, and a fourth dividing unit 138c are provided.

The second average black pixel section length calculator 131b calculates the length of the second black pixel section (second black pixel section length) of the second divided row cut out by the second divided row cutout section 18b. Is determined as a second average black pixel section length. Then, the second average black pixel section length calculation unit 131b calculates
The obtained information on the second average black pixel section length is sent to the second division candidate section selection section 132b.

The second division candidate section selection section 132b
Compares the second average black pixel section length with the second black pixel section length, and determines that the ratio of the second black pixel section length to the second average black pixel section length in the second black pixel section is a sixteenth reference. A second black pixel section larger than the value is selected as a second division candidate section. Then, the second division candidate section selection unit 132b sends information on the selected second division candidate section to the third and fourth position extraction units 134c and 134d. In addition, this first
As the sixth reference value, any suitable value may be set similarly to the above-described eleventh reference value.

Further, the third position extracting section 134c is the most one-side position along the first direction among the positions where black pixels are detected on a scanning line that scans the vicinity of the center of the second division candidate section in the scanning line. Is extracted as the third position. And
The third position extracting unit 134c sends information on the extracted third position to the nearest third position extracting unit 136c.

Further, the nearest third position extracting unit 136c extracts the third position closest to the other side in the first direction among the third positions as the nearest third position. Then, the nearest third position extracting unit 136c sends information on the extracted nearest third position to the third and fourth dividing units 138c and 138d.

Further, the fourth position extracting section 134d detects the position of the black pixel on the scanning line that scans the vicinity of the center of the second division candidate section in the scanning line, and is located closest to the other side in the second direction. Is extracted as the fourth position. And
The fourth position extracting unit 134d sends information on the extracted fourth position to the nearest fourth position extracting unit 136d.

The fourth nearest position extraction unit 136d extracts the fourth position closest to the first position along the second direction from the fourth positions as the nearest fourth position. Then, the nearest fourth position extracting unit 136d sends information on the extracted nearest fourth position to the third and fourth dividing units 138c and 138d.

The third division 138c compares the nearest third position with the nearest fourth position, and determines whether the nearest third position is on the other side of the nearest fourth position or When the nearest third position and the nearest fourth position are located on the same straight line along the first direction, the second division candidate section is divided into two at the center of the second division candidate section. By generating two divided black pixel sections, the line segmentation is corrected.

The fourth division 138d compares the nearest third position with the nearest fourth position, and determines that the nearest third position is one side of the nearest fourth position. By dividing the second division candidate section by a division line passing through the middle between the nearest third position and the nearest fourth position and along the second direction, the one divided area on one side and the other divided area on the other side are divided. The other divided area is generated, one divided area is divided by the scanning line at which the nearest fourth position is obtained to generate two divided black pixel sections, and the other divided area is obtained at the nearest third position. The line segmentation is performed by dividing the scan line into two divided black pixel sections, thereby correcting the line segmentation.

(Operation) Next, with reference to FIG. 13, the operation of the line extracting apparatus 10 of this embodiment, the line extracting method of the present invention, and the first and second line extracting and correcting methods of the present invention will be described. Examples will also be described in detail. FIG.
3 is a flowchart for explaining the overall operation flow of the line segmentation device 10 of the present embodiment.

In FIG. 13, the flow of the line cutout processing when the line direction is known in advance from the step of switching S2 to the end is shown by a solid line (the line direction is the first line).
Two types of processing flows, one for the direction and one for the second direction, are shown. ). The broken line shows the flow of the line cutout process when the line direction is unknown in advance.

(Overall Flow of Operation When Row Direction is Unknown) First, the overall flow of the operation of the line segmenting apparatus 10 when the row direction is unknown in advance will be described. The case where the line direction is unknown is, for example, a case where no information indicating the line direction of the document image is added to the information of the document image 60, or the information indicating the line direction is included in the information of the document image 60. This applies to the case where the assigned information could not be confirmed even if it was assigned.

When the document image 60 is input to the line segmenting device 10, the control unit 28 first activates the line direction confirmation switching unit 12. The activated line direction confirmation switching unit 12 of the activated line direction confirmation switching unit 12 confirms whether information indicating the line direction of the document image 60 is added to the information of the document image 60.
In this case, it is assumed that the document image 60 is not provided with information indicating the row direction. Therefore, the row direction checking unit 12a
Outputs the confirmation result that the line direction is unknown to the switching unit 12b together with the information of the document image 60 (S1 in FIG. 13).

When the switching unit 12b receives the confirmation result that the line direction is unknown from the line direction confirmation unit 12a, the switching unit 12b divides the information of the document image 60 into the first and second direction as shown by the broken line in FIG. Area generating units 14a and 14b
(S2 in FIG. 13).

The first direction divided area generating section 14a divides the document image 60 to generate a first divided area (S3 in FIG. 13). The generated information of the first divided area is sent to the first direction projection distribution creating unit 16a.

Next, the first direction projection distribution creating unit 16a
A first direction projection distribution is created for each first divided region (S4 in FIG. 13). The generated information of the first direction projection distribution is sent to the first divided line cutout unit 18a.

Next, the first divided row cutout section 18a
First divided rows are cut out for each divided area based on the first direction projection distribution (S5 in FIG. 13). Information of the first divided row (for example, the position of the first black pixel section,
(Number, length) are sent to the row direction determination unit 20.

On the other hand, the first direction divided area generation section 14b
The document image 60 is divided to generate a second divided area (S6 in FIG. 13). The information of the generated second divided area is the second divided area.
It is sent to the direction projection distribution creating unit 16b.

Next, the second direction projection distribution creating section 16b
A second direction projection distribution is created for each second divided region (S7 in FIG. 13). The generated information on the second direction projection distribution is sent to the second divided line cutout unit 18b.

Next, the second divided line cutout unit 18b
A second divided row is cut out for each divided area based on the second direction projection distribution (S8 in FIG. 13). Information of the cut-out second divided row (for example, the position of the second black pixel section,
(Number, length) are sent to the row direction determination unit 20.

The line direction determining section 20 determines the line direction of the document image 60 based on the information on the first divided line and the information on the second divided line (S9 in FIG. 13). The determination result in the row direction is sent to the switching output unit 22.

When the determination result of the row direction is the first direction, the switching output section 22 sends the information of the first divided row extracted by the first divided row extracting section 18a to the first correction section 24a. Further, the switching output unit 22 determines that the row direction determination result is the second
In the case of the direction, the information of the second divided line extracted by the second divided line extraction unit 18b is sent to the second correction unit 24b (S10 in FIG. 13).

If the row direction is determined to be in the first direction, the first correction unit 24a corrects the cut-out divided lines as necessary (S11 in FIG. 13). The corrected information of the first divided row is sent to the first divided row connecting unit 26a.

Then, the first divided row connecting section 26a connects the first divided rows to each other and cuts out a row (FIG. 13).
S12).

On the other hand, if the row direction is determined to be in the second direction, the second correction unit 24b corrects the cut-out divided lines as necessary (S13 in FIG. 13). The corrected information on the second divided row is sent to the second divided row connecting unit 26b.

Then, the second divided row connection section 26b connects the second divided rows to each other and cuts out a row (FIG. 13).
S14).

In this way, the line cutting operation when the line direction is unknown in advance is ended.

(Overall Flow of Operation when Row Direction is First Direction) Next, the operation of the line segmenting device 10 when the row direction is known in advance to be the first direction will be described. .

[0214] When the document image 60 is input to the line segmentation device 10, the control unit 28 first activates the line direction confirmation switching unit 12. The activated line direction confirmation switching unit 12 of the activated line direction confirmation switching unit 12 confirms whether information indicating the line direction of the document image 60 is added to the information of the document image 60.
In this case, information indicating that the row direction is the first direction is added to the document image 60. Therefore, the line direction confirmation unit 12a outputs a confirmation result that the line direction is the first direction to the switching unit 12b together with the information of the document image 60 (S1 in FIG. 13).

The switching unit 12b receives the result of the confirmation that the row direction is the first direction from the row direction confirmation unit 12a.
Is indicated by a solid line I on the right side of FIG.
The information of the document image 60 is sent to the first direction divided area generation unit 14a (S2 in FIG. 13).

Next, the first direction divided area generation unit 14a
The document image 60 is divided to generate a first divided area (S3 in FIG. 13). The information of the generated first divided area is the first divided area.
It is sent to the direction projection distribution creating unit 16a.

Next, the first direction projection distribution creating unit 16a
A first direction projection distribution is created for each first divided region (S4 in FIG. 13). The generated information of the first direction projection distribution is sent to the first divided line cutout unit 18a.

Next, the first divided line cutout section 18a
First divided rows are cut out for each divided area based on the first direction projection distribution (S5 in FIG. 13). If the row direction is the first direction, information on the cut-out first divided row (for example, the position, number, and length of the first black pixel section)
Is sent to the first correction unit 24a.

Next, the first correction unit 24a corrects the cut-out divided lines as necessary (S1 in FIG. 13).
1). The corrected information of the first divided row is sent to the first divided row connecting unit 26a.

Next, the first divided row connecting section 26a
Rows are cut out by connecting the divided rows (S12 in FIG. 13).

In this manner, the line cutting operation in the case where it is known in advance that the row direction is the first direction ends.

(Overall Flow of Operation when Row Direction is Second Direction) Next, the operation of the line segmenting device 10 when the row direction is known in advance to be the second direction will be described. .

When the document image 60 is input to the line segmentation device 10, the control unit 28 first activates the line direction confirmation switching unit 12. The activated line direction confirmation switching unit 12 of the activated line direction confirmation switching unit 12 confirms whether information indicating the line direction of the document image 60 is added to the information of the document image 60.
In this case, information indicating that the row direction is the second direction is added to the document image 60. Therefore, the line direction confirmation unit 12a outputs a confirmation result that the line direction is the second direction to the switching unit 12b together with the information of the document image 60 (S1 in FIG. 13).

The switching unit 12b receives the result of the confirmation from the row direction confirmation unit 12a that the row direction is the second direction.
13, the information of the document image 60 is divided by the second direction divided area generation unit 14 as indicated by a solid line II on the left side of the paper of FIG.
b (S2 in FIG. 13).

Next, the second direction divided area generation unit 14b
The document image 60 is divided to generate a second divided area (S6 in FIG. 13). The information of the generated second divided area is the second divided area.
It is sent to the direction projection distribution creating unit 16b.

Next, the second direction projection distribution creating section 16b
A second direction projection distribution is created for each second divided region (S7 in FIG. 13). The generated information on the second direction projection distribution is sent to the second divided line cutout unit 18b.

Next, the second divided line cutout section 18b
A second divided row is cut out for each divided area based on the second direction projection distribution (S8 in FIG. 13). If the row direction is the second direction, information on the cut-out second divided row (for example, the position, number, and length of the second black pixel section)
Is sent to the second correction unit 24b.

Next, the second correction section 24b corrects the cut-out divided lines as necessary (S1 in FIG. 13).
3). The corrected information on the second divided row is sent to the second divided row connecting unit 26b.

Next, in the second connection section 26b, the second divided rows are connected to each other to cut out the rows (S1 in FIG. 13).
4).

In this way, the line cutting operation when the line direction is known in advance to be the second direction ends.

(Regarding Line Cutting Object) Hereinafter, FIG.
The operation of each step when line extraction of the document image 60 shown in FIG. This document image 60
Is a horizontal direction. That is, the document image 60
Contains ten horizontal character strings 62. The character string 62 includes full-width characters and half-width characters, respectively. The character spacing of each character string 62 is not fixed pitch but variable pitch (proportional). In FIG. 14, each character included in each character string is schematically shown by a hatched circle. For example, full-width characters 64a are indicated by a perfect circle, and half-width characters 64a are
b is indicated by a vertically long ellipse.

In FIG. 14, the X axis is taken in the horizontal direction on the paper surface and the Y axis is taken in the vertical direction on the paper surface. Hereinafter, the X-axis direction is also referred to as a horizontal direction. Further, the Y-axis direction is also referred to as a vertical direction.

In the document image 60, a character area 66 is set in the character area setting section 32 shown in FIG. Then, the document image 60 in which the character area 66 is set
Is input to the line segmentation device 10.

(Regarding Document Image Division Processing) Next, an example of document image division processing for the document image 60 will be described.

If the line direction of the document image 60 is unknown or the line direction is the first direction, the control unit 28 activates the first direction divided area generation unit 14a. As shown in FIG. 15A, the activated first direction divided area generation unit 14a generates the document image 60 by using the dividing line AA and the dividing line BB along the second direction (vertical direction). Character area 66
Is divided (S3 in FIG. 13). Further, by dividing the character area 66, each character string (line) 6 in the character area 66 is divided.
2 is also divided respectively. When the character area 66 is divided, the document image 60 is also divided as a result.

As shown in FIG. 15A, in this embodiment, in this embodiment, three first directions arranged along the first direction orthogonal to the second direction, that is, along the horizontal direction, are obtained. A divided area (horizontal divided area) 68 is generated. Here, as the horizontal division area 68, the first division area 68a, the second division area 68b, and the third division area 6
8c respectively.

If the row direction is unknown or the row direction is the second direction, the control unit 28 activates the second direction divided area generation unit 14b. As shown in FIG. 15B, the activated second direction divided area generation unit 14b sets the dividing line along the first direction (horizontal direction) to C-C and the dividing line D-
The document area 66 of the document image 60 is divided by D and the dividing line EE (S6 in FIG. 13).

In this embodiment, as shown in FIG. 15B, this division makes it possible to form four second directions orthogonal to the first direction, that is, four second directions arranged along the vertical direction. A divided area (vertical divided area) 70 is generated.
Here, the fourth divided area 70 is used as the vertical divided area 70.
a, the fifth divided area 70b, the sixth divided area 70c, and the seventh divided area 70d are respectively generated.

In this embodiment, the character area 66 is horizontally divided into three divided areas, but the number of divisions can be set to any suitable value (natural number). For example, the number of divisions may be set as appropriate according to conditions such as the resolution of the character area of the document image and the area of the character area.

For example, when the width of the character area 66 along the horizontal direction (X-axis direction, horizontal direction) is defined as Wa and a horizontal divided area whose maximum width is equal to or smaller than Wth is generated, the number of divisions n in the horizontal direction is n.
May be obtained by the following (1).

N = (Wa + Wth) / Wth (1) In the above equation (1), the quotient of the division is represented by an integer, and the remainder of the division is discarded.

When the character area 66 is equally divided into n, the width of each horizontal division area is preferably set to Wa / n pixels (dots).

Also, for example, the height along the vertical direction (Y-axis direction, vertical direction) of the character area is defined as Ha, and the maximum height is defined as H.
When generating a vertical division area equal to or less than th, the number m of divisions in the vertical direction may be obtained by the following equation (2).

M = (Ha + Hth) / Hth (2) In the above equation (2), the remainder of the division is rounded down, and the quotient of the division is represented by an integer.

When the character area 66 is equally divided into m, the height of each vertical division area is preferably set to Ha / m pixels (dots).

In the present invention, the character area 66 does not necessarily need to be divided evenly. In the present invention,
The document image 60 may be divided without setting the character area 66.

(Regarding Projection Distribution Creation Processing) Next, the projection distribution creation processing will be described in more detail. When the row direction is unknown or the row direction is the first direction, the control unit 28 activates the first direction projection distribution creation unit 16a. The activated first direction projection distribution creating unit 16a
For each of the first direction divided regions (horizontal divided regions) 68, scanning is performed by scanning the first direction divided region with the first direction (horizontal direction) as the main scanning direction and the second direction (vertical direction) as the sub-scanning direction. The first direction projection distribution (horizontal projection distribution) 72 is created by detecting the cumulative number of black pixels for each line (S4 in FIG. 13).

In this embodiment, the first direction projection distribution 72 is created for the first divided area 68a of the first direction divided area 68. Here, FIG.
The first divided area 68a and its first direction projection distribution 72 are shown. The first direction projection distribution 72 represents the cumulative number of black pixels for each scanning line in a histogram. Therefore, the height of the projection distribution relatively indicates the cumulative number of black pixels.

The first direction projection distribution 72 is obtained by scanning the document image 60 along the row direction. For this reason, the first direction projection distribution 72 includes a plurality of mutually independent projection blocks 74. The first direction projection distribution 72 is stored in the internal memory of the first direction projection distribution creating unit 16a.

When the row direction is unknown or when the row direction is the second direction, the control unit 28 activates the second direction projection distribution creating unit 14b. The activated second direction projection distribution creating unit 16b sets the second direction (vertical direction) to the main scanning direction and sets the first direction (horizontal direction) to the sub-scanning for each of the second direction divided areas (vertical divided areas) 70. The second direction projection distribution (vertical projection distribution) 7 is obtained by scanning the second direction divided area as the direction and detecting the cumulative number of black pixels for each scanning line.
8 (S7 in FIG. 13).

In this embodiment, the second direction projection distribution 78 is created for the fourth divided area 70a of the second direction divided area 70. Here, FIG.
The first divided region 70a and its second direction projection distribution 78 are shown. The second direction projection distribution 78 represents the cumulative number of black pixels for each scanning line in a histogram. Therefore, the height of the projection distribution relatively indicates the cumulative number of black pixels.

The second direction projection distribution 78 is obtained by scanning the document image 60 in a direction crossing the row direction. Therefore, the second direction projection distribution 78 is an irregularly continuous projection. The data of the second direction projection distribution 78 is stored in a memory inside the second direction projection distribution creating unit 16b.

(Regarding the process of extracting a divided line) Next, the process of extracting a divided line will be described in more detail.

When the row direction is unknown or when the row direction is the first direction, the control unit 28 activates the first divided row cutout unit 18a. Activated first divided line cutout unit 1
8a, a section in which the cumulative number of black pixels equal to or more than a first reference value (for example, one pixel) in each of the first direction projection distributions 72 is continuous in the second direction (vertical direction) is defined as a first black pixel section. . Then, the first divided line cutout unit 18a performs divided line cutout with the scanning regions in which the first black pixel sections are obtained as first divided lines (S in FIG. 13).
5).

As a result, each sub-scanning section obtained from each projection block 74 in the first direction projection distribution 72 corresponds to a divided line 76 generated by dividing each character string 62 by the dividing line AA. I do. In FIG. 16A, each of the 10 divided lines 76 cut out is defined by a rectangular frame 76.
Are shown schematically. The extraction of the divided rows is performed for all the first divided areas 68.

When the row direction is unknown or the row direction is the second direction, the control unit 28 activates the second divided row cutout unit 18b. The activated second divided line cutout unit 18b is a section in which the cumulative number of black pixels equal to or more than the second reference value (for example, one pixel) is continuous along the first direction (horizontal direction) in each of the second direction projection distributions 78. Are each a second black pixel section. Then, the second divided line cutout unit 18b
Performs segmented line segmentation using the scanning area in which each of the second black pixel sections is obtained as a second segmented row (FIG. 1).
3 S8).

The second direction projection distribution 78 is interrupted at portions indicated by reference numerals a, b, c, and d in FIG. 16B. Then, in FIG. 16B,
Each divided line 80 generated by dividing the character string 62 included in the fourth divided region 70a at each part where the projection of the second direction projection distribution 78 is interrupted is schematically shown by a rectangular frame 80. Each divided line 80 is arranged in the first direction (horizontal direction) with each other, and includes a part of each of the plurality of character strings 62. FIG. 16B shows an example in which five divided rows 80 are generated.

Hereinafter, the divided row 76 in the first divided area 68a
Is referred to as a horizontal divided row 76, and the divided row 78 in the fourth divided area 70a is referred to as a vertical divided row 80.

In this embodiment, FIG.
Steps S3 to S5 and steps S6 to S8 may be processed in parallel with each other.

(Regarding Operation of Line Direction Judgment Unit) Next, an example of operation of the line direction judgment unit when the line direction of the document image 60 is unknown in advance will be described.

The operation (S9 in FIG. 13) of the row direction determination in the row direction determination section 20 will be described with reference to FIG. FIG. 17 is a flowchart for explaining the operation of the row direction determination unit 20. Hereinafter, the first part of the row direction determination unit 20 will be described.
Determination unit 201, second determination unit 202, and third determination unit 20
Operation 3 will be described sequentially.

(Regarding the Operation of the First Judgment Unit) First, the first
As an example of the operation of the determination unit 201, an operation when the document image 60 including a plurality of lines is determined as illustrated in FIG. 14 will be described.

The control unit 28 first activates the first determination unit 201 of the row direction determination unit 20. Then, the control unit 28
To the activated first determination unit 201, the first divided line cutout unit 1
8a and the data of the first projection distribution and the second projection distribution, respectively, are sent from the second divided row cutout unit 18b.

In the first direction projection distribution creating section 16a, the first direction projection distribution 72
It was created. Therefore, the first divided region number counting unit 36a of the first determination unit 201 counts the first divided region number n, which is the total number of the generated divided regions 68, as “3”.

The first direction projection distribution 72 of each divided area
As described above, each of the ten projection blocks 7
4 is included. Each projection block 74 is a first black pixel section in which the projection of the accumulated black pixels of the first reference value (here, one pixel) or more in the first direction projection distribution 72 is continuous in the first direction. Therefore, the first black pixel section number counting unit 38a of the first determination unit 201 sets the first black pixel section number Nh, which is the sum of the number of black pixel sections in the first direction projection distribution over all the divided regions, to " 30 (10 × 3) ”.

In the second direction projection distribution creating section 16b, the second direction projection distribution 78 is created for all the divided areas 70. Therefore, the second divided region number counting unit 36b of the first determination unit 201 calculates the second divided region number that is the total number of divided regions.
The number m of divided areas is counted as “4”.

In addition, the second direction projection distribution 78 of each divided area
Each include five second black pixel sections. For example, as described above, the second direction projection distribution 78 of the fourth divided region 70a is interrupted at four positions a, b, c, and d. Each of the five projection distributions interrupted at the four locations is the second projection distribution 78 in the second direction projection distribution 78.
This is a second black pixel section in which the projection of the cumulative number of black pixels equal to or more than the reference value (here, one pixel) is continuous in the first direction (lateral direction).
Here, it is assumed that the same number of black pixel sections are included in the second direction projection distribution of the other second divided areas. Therefore, the second black pixel section number counting section 38b of the first determination section 201 sets the second black pixel section number Mv, which is the sum of the number of each second black pixel section over all divided regions, to "20".
(5 × 4) ”.

The first divided area number counting section 36a,
The order of the processes of the second divided region number counting section 36b, the first black pixel section number counting section 38a, and the second black pixel section number counting section 38b is not limited.

Next, the section number division number comparison judgment section 40 of the first judgment section 201 compares the first black pixel section number Nh with the first division area number n, and compares the second black pixel section number Mv with the second black pixel section number Mv. Compare the number m of the second divided areas (S1 and S1 in FIG. 17).
2).

In this embodiment, first, n = Nh and m ≠
It is determined whether it is Mv (S1 in FIG. 17). If n = Nh and m ≠ Mv (Y), the document image 60 is determined to be horizontally written. On the other hand, n =
If Nh and m ≠ Mv are not satisfied (N), the process proceeds to the next step S2.

In step S2, n ≠ Nh and m
= Mv (S2 in FIG. 17). If n ≠ Nh and m = Mv (Y), the document image 60 is determined to be vertically written. On the other hand, n ≠
If Nh and m = Mv are not satisfied (N), the process proceeds to the next steps S3 and S4.

Here, the first black pixel section number Nh (10)
Does not match the first divided area number n (1), and does not match the second black pixel section number Mv (5) and the second divided area number m (1). Therefore, in this case, the first determination unit 201 cannot determine the row direction.

Then, the control unit 28 executes the next step S3
In order to proceed to steps S4 and S4, the second determination unit 202 of the row direction determination unit 20 makes a determination.

(Example of Operation of First Judgment Unit in Case of One Line of Character String) As described above, the document image 60 subjected to the judgment of the line direction in this embodiment includes a plurality of lines. Have been. Therefore, the first determination unit 201 does not determine the line direction of the document image 60.

In order to better understand the operation of the first determination unit 201, an example of the operation of determining the line direction of a document image including only one line of character string will be described.
Here, the character area 50 of the document image is shown in FIG.
Shows an example in which the horizontal character string 52 includes only one line. In FIG. 18, the area of the character string 52 is indicated by hatching for convenience. Hereinafter, in this embodiment, the horizontal direction will be described as a first direction, and the vertical direction will be described as a second direction.

First, as shown in FIG. 18B, in the above-described first direction divided area generating section 14a, the character area 50 is divided into three parts by a vertical dividing line AA and a dividing line BB. Will be described. As a result of this division, a first divided region 50a, a second divided region 50b, and a third divided region 50c are sequentially arranged in the horizontal direction.
As a result of this division, the character string 52 is also changed to the first divided line 52.
a, a second divided row 52b and a third divided row 52c. The first divided row 52a is included in the first divided area 50a. In addition, the second divided row 52b includes the second divided area 5
0b. The third divided row 52c is included in the third divided area 50c.

Then, the above-described first direction projection distribution creating section 1
In 6a, a horizontal projection distribution (first direction projection distribution) is applied to two of the three divided regions, for example, a first divided region 50a and a second divided region 50b.
Is created respectively.

Here, FIG. 18C shows the first divided area 50a. The first divided area 50a includes only one first divided row 52a. Then, on the right side of the first divided region 50a, a first direction projection distribution 54a obtained by scanning the first divided region 50a in the horizontal direction is shown. This first
The direction projection distribution 54a is the one corresponding to the first divided row 52a.
One block-shaped projection. This block-shaped projection corresponds to one black pixel section 56a. Therefore, one black pixel section 56a is obtained for one divided region 50a.

FIG. 18D shows the second divided area 5.
0b is indicated. The second divided area 50b includes only one second divided row 52b. The first direction projection distribution 54b obtained by scanning the second divided region 50b in the horizontal direction is shown on the right side of the second divided region 50b. The first direction projection distribution 54b is one block-shaped projection corresponding to the second divided row 52b. This block-shaped projection is 1
One black pixel section 56b. Therefore, one black pixel section 56b is obtained for one divided region 50b.

Then, as described above, the number of divided areas is 3
One. Therefore, the first divided area number counting section 36a
Sets the first number of divided areas, which is the total number of divided areas, to "3".
Is counted.

Further, as described above, the first divided area 50a
The number of black pixel sections 56a of the first direction projection distribution 54a is “1”. Further, the number of black pixel sections 56b of the first direction projection distribution 54b of the second divided area 50b is also “1”.
Further, the number of black pixel sections 56c of the first direction projection distribution 54c of the third divided area 50c is also “1”.

Therefore, the first black pixel section number counting section 38a
Counts the number of first black pixel sections, which is the sum of the number of black pixel sections in the first direction projection distribution over all divided regions, as “3” (= 1 + 1 + 1).

Next, as shown in FIG. 19A, in the above-described second direction divided area generating section 12b, the character area 50 shown in FIG. The case where the image is divided into three by the dividing line DD will be described. As a result of this division, the fourth divided region 50d, the fifth divided region 50e, and the sixth divided region 5 sequentially arranged in the vertical direction
Three divided areas of 0f are generated. And the character string 5
2 is included in the fifth divided region 50e.

The first direction projection distribution creating unit 1 described above
In step 6a, a vertical projection distribution (second direction projection distribution) is created for each of the three divided regions.

Here, FIG. 19B shows a fourth divided region 50d. The character string 52 is not included in the fourth divided area 50d. Therefore, even if the fourth divided region 50d is scanned in the vertical direction, a projected distribution cannot be obtained. Therefore, a black pixel section cannot be obtained from the fourth divided area 50d. Similarly, a black pixel section cannot be obtained from the sixth divided area 50f.

FIG. 19C shows the fifth divided area 5.
0e. The character string 52 is directly included in the fifth divided area 50e. The second direction projection distribution 54e obtained by scanning the fifth divided area 50e in the vertical direction is shown below the fifth divided area 50e. This second direction projection distribution 5
4e is one continuous projection. This projection corresponds to one black pixel section 56e. Therefore, one black pixel section 56e is obtained for one divided region 50e.

Then, the second divided area number counting section 36b
Sets the number of second divided areas, which is the total number of divided areas, to "3"
Is counted.

As described above, the number of black pixel sections in the fourth and sixth divided areas 50d and 50f is "0". The number of black pixel sections 56e in the second direction projection distribution 54e of the fifth divided region 50e is “1”.

Therefore, the second black pixel section number counting section 38b
Counts the number of second black pixel sections, which is the sum of the number of black pixel sections in the second direction projection distribution over all the divided regions, as “1” (= 0 + 1 + 0).

Next, the section number division number comparison judgment section 40 judges the row direction. Here, the first black pixel section number (3) matches the first divided area number (3), and the second black pixel section number (3) matches the second black pixel section number (3).
Since the number of black pixel sections (1) does not match the number of second divided areas (3), the row direction is determined to be the first direction, that is, the horizontal direction.

As described above, when only one line of a character string is included in the character region of the document image, one divided line is included in each divided region divided by the dividing line orthogonal to the line direction. It is. As a result, in this case, the number of divided areas matches the total number of divided rows included in each divided area. On the other hand, when the character area is divided by the dividing line along the line direction, the number of divided areas does not match the total number of divided lines. For example, even if a plurality of divided areas are generated, the total number of divided rows remains one. Therefore, when the character string has only one line, if the number of divided regions and the total number of divided lines are compared, it is not necessary to compare the variation of the black pixel section length and the average value of the white pixel section length.
The row direction can be determined.

Therefore, when the character string in the document image is only one line, the first judgment unit 201 compares the number of divided areas with the total number of divided lines, thereby making the second judgment unit 20.
2, it is not necessary to compare the variation of the black pixel section length as described later, or to compare the average value of the white pixel section length as described later in the third determination unit 203, thereby improving the determination accuracy in the row direction. be able to.

In the operation example described above, the two first
Although the projection distribution is created for each of the direction divided areas and the two second direction divided areas, for example, the projection distribution may be created for all divided areas. Further, the projection distribution may be created only for one first direction divided region and one second direction divided region.

Further, for example, the second divided area 50b and the fifth divided area 50b
When the projection distribution is created only for the divided region 50e, both the first divided region number and the second divided region number are “1”. Then, both the first black pixel section number and the second black pixel section number are “1”. In this case, the first
Since the number of divided areas “1” is equal to the number of first black pixel sections “1”, and the number of second divided areas “1” is equal to the number of second black pixel sections “1”, the number of divided sections is compared. The determination condition in the determination unit 40 is not satisfied. Therefore, in this case,
The section number division number comparison determination unit 40 does not determine the row direction. In that case, the second determination unit 202 attempts to determine the row direction.

(Regarding the Operation of the Second Judgment Unit)
The operation of the determination unit will be described.

Next, the control unit 28 activates the second judgment unit 202 of the judgment unit 20. Then, the control unit 28 sends the data of the first projection distribution and the data of the second projection distribution to the activated second determination unit 202 from the first divided line cutout unit 18a and the second divided line cutout unit 18b, respectively.

(Regarding the Operation of the First Variation Calculation Unit) First, the operation of the first variation calculation unit 42a of the second determination unit 202 will be described with reference to FIG. FIG.
(A) of FIG. 0 is a diagram provided for explanation of the first variation calculation unit 42a, and the first divided area 6 shown in (A) of FIG.
8A shows a portion of the portion 8a and a portion of the first direction projection 72 corresponding to the portion.

In FIG. 20A, the divided rows 76 in the first divided area 68a are arranged in order from the upper side of the drawing sheet into a first divided row 76a, a second divided row 76b, a third divided row 76c,.
-The tenth divided row 76j is used. Also, the first divided row 76
a, second divided row 76b, third divided row 76c... tenth
The projection blocks 74 obtained by scanning the divided rows 76j are referred to as first projection blocks 74a, second projection blocks 74b, third projection blocks 74c,..., Tenth projection blocks 74j. In FIG. 20A, illustration of the fourth to ninth projection blocks is omitted.

The first variation calculator 42a first determines that the cumulative number of black pixels equal to or greater than the first reference value (for example, one pixel) in the cumulative number of black pixels forming the first direction projection distribution 72 When it is continuous along (vertical direction), the continuous section 82 is defined as a first black pixel section, and the length of the first black pixel section 82 (first black pixel section length) is obtained. Here, the first projection block 74a and the second projection block 7
4b, third projection block 74c... The first black pixel section length of each black pixel section 82 corresponding to the tenth projection block 74j is H (1), H (2), H (3).
Determined as (10).

The data of the first black pixel section length is obtained, for example, by obtaining the XY coordinates of the position where the end of each projection block 74 is obtained in the document image 60, and calculating the XY coordinates of the end of each projection block 74 in the sub-scanning direction. XY coordinate distance (in this case, Y
(Difference in coordinate values). Further, the data of the first black pixel section length may be represented by, for example, the number of pixels between both ends of each projection block 74 along the sub-scanning direction.

When the first reference value is set to one pixel, each first black pixel section length corresponds to each projection block 7.
4 in the sub-scanning direction. Each projection block 74
In the sub-scanning direction respectively correspond to the height of each divided row 76. On the other hand, when the first reference value is a value of two or more pixels (for example, several pixels), each first black pixel section length is longer than the length of each projection block 74 in the sub-scanning direction. Also tend to be slightly shorter. However, also in this case, the length of the black pixel section substantially corresponds to the length of the projection block in the sub-scanning direction and the height of the divided row 76. The first reference value can take any suitable number of pixels (natural number).

When the first black pixel section lengths have already been obtained, for example, the first direction projection distribution creating section 16a sets the first black pixel section lengths.
When the black pixel section length has already been obtained, it is not necessary for the first variation calculation section 42a to newly obtain each first black pixel section length.

Next, the first variation calculating section 42a calculates the first
The variation of the black pixel section length is calculated (S3 in FIG. 17).
In calculating the variation of the first black pixel section length, the first variation calculation unit 42a first calculates the arithmetic mean of the black pixel section length of each black pixel section 82. The arithmetic mean value Hh in the black pixel section is given by the following equation (3).

Hh = {H (1) + H (2) + H (3) +... + H (10)} / Nh (3) where Nh represents the number of black pixel sections 82. In the embodiment, Nh = 10. Further, a plurality of divided areas 68
In the case where the first direction projection distribution 72 is created, the total number of black pixel sections 82 included in each of the divided regions 68 in which the first direction projection distribution 72 is created is Nh.
For example, when the first direction projection distribution 72 is created for two divided regions 68, the total number of black pixel sections 82 is Nh = 20.

Next, the first variation calculator 42a calculates a value obtained by dividing the sum of the difference between each black pixel section length and the arithmetic mean value Hh by the number of black pixel sections as the first pixel section length variation Eh. It is calculated (S4 in FIG. 17).

The variation Eh of the first black pixel section length is given by the following equation (4).

Eh = ｛Σ | H (i) −Hm |｝ / Nh (4) where “Σ” in the above equation (4) represents the sum when i = 1 to Nh. .

In this embodiment, the variation Eh of the first black pixel section length is given by the above equation (4).
The variation in the first black pixel section length is not limited to this. In the present invention, for example, a variance or a standard deviation may be obtained as the variation of the first black pixel section length.

(Regarding the Operation of the Second Variation Calculation Unit) Next, referring to FIG.
The operation of the variation calculator 42b will be described. FIG.
(B) of FIG. 16 is a diagram provided for explanation of the second variation calculation unit 42b, and the fourth divided area 70 shown in (B) of FIG.
a and the second direction projection 78 are shown.

In FIG. 20 (B), the divided rows 80 in the fourth divided area 70a are arranged in order from the left side of the page, as a first divided row 80a, a second divided row 80b, a third divided row 80c, and a fourth divided row 80c. A divided row 80d and a fifth divided amount 80e are set. Also,
First divided row 80a, second divided row 80b, third divided row 80
c, a projection 78 obtained by scanning the fourth divided row 80d and the fifth divided row 80e, respectively.
8a, second projection 78b, third projection 78c, fourth projection 78
d and the fifth projection 78e.

[0311] Then, the second variation calculation unit 42b first determines, among the cumulative black pixels forming the second direction projection distribution 78, the cumulative number of black pixels equal to or more than the second reference value (for example, one pixel) in the first direction. When it is continuous along the (horizontal direction), the continuous section 84 is defined as a second black pixel section, and the length of the second black pixel section 84 (second black pixel section length) is obtained. Here, the first projection 78a, the second projection 78b, and the third projection 7
8c, the second black pixel section length of each black pixel section 84 corresponding to the fourth projection 78d and the fifth projection 78e is W
(1), W (2), W (3), W (4) and W (5)
Asking.

The data of the second black pixel section length is obtained, for example, by calculating the XY coordinates of the position of the end of each projection 78,
8 may be expressed as the distance between the XY coordinates at both ends along the sub-scanning direction (in this case, the difference between the X coordinate values). Also, the second
The data of the black pixel section length may be displayed by, for example, the number of pixels between both ends of each projection 78 along the sub-scanning direction.

If the length of each second black pixel section has already been determined, for example, the second direction projection distribution creating unit 16b may use
If the black pixel section length has already been determined, it is not necessary for the second variation calculation section 42b to determine each second black pixel section length again.

Next, the second variation calculator 42b calculates the second
The variation of the black pixel section length is calculated (S5 in FIG. 17).
In calculating the variation of the second black pixel section length, the second variation calculation unit 42b first calculates the arithmetic mean of the black pixel section length of each black pixel section 84. The arithmetic mean value Wv of the black pixel section is given by the following equation (5).

Wv = {W (1) + W (2) + W (3) + W (4) + W (5)} / Nw (5) where Nw represents the number of black pixel sections 84. In the embodiment, Nw = 5. When the second direction projection distribution 78 is created for a plurality of divided areas 70, the first
The total number of black pixel sections 84 included in each of the divided areas 70 for which the directional projection distribution 78 is created is Nw.

Next, the second variation calculation section 42b calculates a value obtained by dividing the sum of the difference between each second black pixel section length and the arithmetic mean value Wv by the number of black pixel sections, and calculates the variation of the second pixel section length. It is calculated as Ev (S6 in FIG. 17).

The variation Ev of the second black pixel section length is given by the following equation (6).

Ev = ｛Σ | W (j) −Wv |｝ / Nw (6) where “Σ” in the above equation (6) represents the sum when j = 1 to Nw. .

In this embodiment, the variation Ev of the second black pixel section length is given by the above equation (6).
The variation of the second black pixel section length is not limited to this. In the present invention, for example, a variance or a standard deviation may be obtained as the variation of the second black pixel section length.

Further, in this embodiment, the processing for obtaining the variation in the first and second black pixel sections is performed in the order of steps S3 to S6 in FIG. The processing in step S5 and the processing in steps S4 and S6 can be performed independently and independently. Therefore, in the present invention, S3
The order of processing is not limited to steps S6 to S6, except that steps S3 and S4 are sequentially performed and steps S5 and S6 are sequentially performed.
For example, the step of S3 may be performed after the step of S5 is performed. Further, for example, after the processes of steps S5 and S6 are sequentially performed, the processes of steps S3 and S4 may be sequentially performed.

(Operation of Variation Comparison Determining Unit) Next, the operation of the variation comparison determining unit 44 of the second determining unit 20 will be described. The variation comparison determination unit 44 compares the variation Eh of the first black pixel section length with the variation Ev of the second black pixel section length to determine the variation Eh of the first black pixel section length.
If the ratio of the variation of the black pixel section length to the second black pixel section length Ev is equal to or greater than the third reference value P, the scanning direction in which the black pixel section length of the smaller variation is obtained is determined as the row direction (S7 in FIG. 17).
And S8). The value of the reference value P can be set to any suitable value. When P = 1, Eh
This is equivalent to comparing the magnitudes of Ev and Ev.

In this embodiment, first, the following (7)
It is determined whether the formula is satisfied (S7 in FIG. 17). When the following expression (7) is satisfied, the document image 60 is determined to be horizontally written. In contrast, the following (7)
If the expression is not satisfied, the process proceeds to the next step S8.

Eh · P <Ev (7) In step S8, it is calculated whether the following equation (8) is satisfied (S8 in FIG. 17). When the following expression (8) is satisfied, the document image 60 is determined to be vertically written. On the other hand, if the following equation (8) is not satisfied, the process proceeds to the next steps S9 and S10.

Ev · P <Eh (8) Here, since the document image 60 is written horizontally, the variation Ev of the second black pixel section length is larger than the variation Eh of the first black pixel section length. Become. As a result, in this case, the above equation (7) is usually satisfied.

By the way, as shown in FIG. 20B, when the character spacing in each line is a fixed pitch, the first
Since both the variation in the black pixel section length and the variation in the second black pixel section length are small, the above equations (7) and (8) are used.
Neither of the expressions may be satisfied. In that case,
The control unit 28 proceeds to steps S9 and S10,
The second determination unit 202 of the row direction determination unit 20 makes a determination.

(Regarding the operation of the third determination unit) Next, the control unit 28 activates the third determination unit 203 of the row direction determination unit 20 and sends the first divided line cutout unit to the activated third determination unit 203. The data of the first projection distribution and the data of the second projection distribution are sent from the segmented line extraction unit 18a and the second divided line segmentation unit 18b, respectively.

(Regarding the Operation of the First Average Value Calculating Unit) The first average value calculating unit 46a
In step 2, the length (first white pixel section length) of the first white pixel section 86 between adjacent first black pixel sections 82 is determined.

In this case, as shown in FIG.
The section length of the first white pixel section between the black pixel section of the first projection block 74a and the black pixel section of the second projection block 74b is determined as Sh (1). Also, the second projection block 7
The section length of the first white pixel section between 4b and the third projection block 74c is obtained as Sh (2). Similarly, the section lengths Sh (3), Sh (4), and Sh of other white pixel sections
(5), Sh (6), Sh (17th), Sh (8) and Sh (9) are obtained.

When the first white pixel section length has already been obtained, for example, when the first direction projection distribution creation section 16a has already obtained each white pixel section length, the first average value calculation section It is not necessary to newly calculate each white pixel section length in 46a.

Next, the first average value output unit 46a calculates the average value Shm of the first white pixel section length (S9 in FIG. 17).
In calculating the average value of the first white pixel section length, the first average value calculation unit 46a calculates the arithmetic mean of the lengths of the first white pixel sections 86 as the first average value. First average value S
hm is given by the following equation (9).

Shm = {Sh (1) + Sh (2) + Sh (3) +... + Sh (9)} / (Nh−n) (9) where Nh−n is the white pixel section 86 Represents the sum of the numbers.
In this embodiment, Nh-n = 9. That is, N
h represents the number of the first black pixel sections 82 as described above, and n
Represents the number of divided areas 68 for which the first direction projection distribution 72 has been created (n = 1 in this embodiment). Also, for example,
In the case where the first direction projection distribution 72 is created for the two divided regions 68 and the arithmetic mean value of the white pixel section combining the two divided regions 68 is calculated, n = 2. That is, Nh-n is a value obtained by adding a value obtained by subtracting 1 from the number of the first black pixel sections 82 in each divided area 68.

In the case where there is only one white pixel section between black pixel sections in the first direction projection distribution, the section length of the white pixel section is set as a first average value.

(Regarding the Operation of the Second Average Value Calculation Unit) The second average value calculation unit 46b obtains the interval between adjacent second black pixel sections as the second white pixel section length.

In this case, as shown in FIG.
The second white pixel section length of the portion indicated by a in FIG. 20B between the first projection 78a and the second projection 78b is Sw.
Determined as (1). In addition, the length of the second white pixel section between the second projection 78b and the third projection 78c, which is indicated by b in FIG. 20B, is determined as Sw (2). Also,
The partial second white pixel section length indicated by c in FIG. 20B between the third projection 78c and the fourth projection 78d is Sw.
Determined as (3). In addition, the partial second white pixel section length indicated by d in FIG. 20B between the fourth projection 78d and the fifth projection 78e is obtained as Sw (4). These second white pixel section lengths Sw (1) to Sw (4) are shown in FIG.
In (B), each pixel is about one pixel.

If the length of each second white pixel section has already been obtained, for example, the second direction projection distribution creating unit 16b may use
If the white pixel section length has already been obtained, it is not necessary to newly calculate each second white pixel section length in the second average value output section 46b.

Next, the second average value output section 46b calculates the average value Swm of the second white pixel section length (S1 in FIG. 17).
0). In calculating the average value of the second white pixel section length,
The second average value calculator 46b calculates an arithmetic average of the white pixel section length of each second white pixel section 88 as a second average value.
The second average value Swm is given by the following equation (10).

Swm = {Sw (1) + Sw (2) + Sw (3) + Sw (4)} / (Nw-m) (10) where Nw-m is the number of the second white pixel sections 88. In this embodiment, Nw−m = 4. That is, Nw
Represents the number of the second black pixel sections 84 as described above, and m
Is the second direction divided area 7 in which the second direction projection distribution 78 is obtained.
Represents the number 0 (m = 1 in this embodiment).

In the case where there is only one white pixel section between black pixel sections in the second direction projection distribution, the section length of the white pixel section is set to a second average value.

In this embodiment, an example in which the process proceeds to step S8 after step S7 has been described.
After step S8, the process may proceed to step S7.

(Regarding the operation of the average value comparing and judging unit) Next, the operation of the average value comparing and judging unit 48 will be described. The average value comparison determination unit 48 compares the first average value Shm and the second average value Swm, and determines the scanning direction of the projection distribution obtained with the white pixel section length of the larger average value as the row direction ( S11 in FIG. 17).

In this embodiment, Shm ≧ S
If wm is satisfied (Y), it is determined that the document image 60 is written horizontally. On the other hand, if Shm ≧ Swm is not satisfied (N), the document image 60 is determined to be vertically written.

In this case, Shm> Swm. Therefore, the average value comparison determination unit 48 determines the scanning direction in which the first direction projection distribution 72 is obtained, that is, the horizontal direction as the row direction. Therefore, the document image 60 is determined to be horizontally written.

In this embodiment, the row direction is determined by comparing the first average value Shm and the second average value Swm in the average comparison / determination section. However, the length of the first white pixel section and the second In determining the row direction by using the length of the white pixel section, it is not necessary to limit to the case of using the average value. For example, the row direction may be determined by comparing specific white pixel sections, for example, by comparing the maximum first white pixel section length with the maximum second white pixel section length. Also, for example, a specific first
Alternatively, the row direction may be determined by comparing the lengths of the second white pixel sections.

(Example of Operation of Integration Correction) Next, an example of a correction method in the case where one divided row is erroneously cut out as two divided rows will be described.

FIGS. 22A and 22B show the divided rows to be corrected. FIG. 22A shows an example of a horizontally divided line 150. This division row 150 is
It contains three characters, "two", "kichi" and "go".
Each of these three characters has a vertically separated radical. For this reason, in the process of separating the divided lines, the divided line 150 to be cut out as one divided line is erroneously divided into two divided lines, that is, an upper divided line 150a and a lower divided line 1a.
It has been cut out as 50b.

FIG. 22 (B) shows an example of a vertically divided line 152. This division line 152 is "eight",
It contains three characters, “north” and “tag”. And
Each of these three characters has a radically separated left and right radical. For this reason, in the process of separating the divided lines, the divided line 152 to be cut out as one divided line is erroneously divided into two divided lines, that is, the right divided line 152a and the left divided line 152b. It has been cut out.

Here, the first direction is defined as a horizontal direction (horizontal direction on the paper, a direction along the X-axis), and the second direction is defined as a vertical direction (vertical direction on the paper, a direction along the Y-axis). Therefore,
The process of integrating and correcting the horizontally written divided row 150 shown in FIG. 22A is performed by the first integrated correcting unit 90a. The process of integrating and correcting the vertically divided lines shown in FIG. 22B is performed by the second integrated correcting unit 90b. If the first direction is changed to the vertical direction and the second direction is changed to the horizontal direction, the processing contents of the first and second integrated correction units 90a and 90b are also changed.

(Regarding the Correction of Integration of Horizontal Split Lines) First,
With reference to FIG. 21, an example of the method of integrating and correcting the horizontally divided rows 150 shown in FIG. FIG.
9 is a flowchart for explaining a method of integrating and correcting horizontal divided rows.

In general, the height of a horizontally divided row that is erroneously cut into two is lower than that of another appropriately divided row, and is about half that of the other divided rows. Furthermore, the line interval between the two erroneously extracted divided lines is generally smaller than the line interval between the other divided lines. Therefore, if attention is paid to the height or width and the line interval of the cut-out divided lines, a divided line that is supposed to be cut out as one divided line by mistake is found as two divided lines, and the divided line is divided. Modification of line segmentation is possible.

Therefore, in the present embodiment, when performing the integrated correction of the horizontal divided rows, first, in the first average black pixel section length calculating section 120a, the first divided row extracted by the first divided row extracting section 18a The arithmetic mean of the length of the first black pixel section (first black pixel section length) of the row is obtained as the first average black pixel section length (S1 in FIG. 21). When the first divided row is a horizontal divided row, the first black pixel section length of the first divided row indicates the height of the horizontal divided row. And in this case,
The first average section length Hh represents an average value of the heights of the horizontal divided rows.

Next, in order to sequentially check whether or not each horizontally divided row is erroneously cut out, a parameter i for conveniently giving the serial number of the horizontally divided row is supplied to the first candidate section selecting unit 122a. prepare. Then, 1 is substituted for the parameter i (S2 in FIG. 21).

The maximum value of the parameter i is the total number Nh of horizontal divided rows.

Next, the first candidate section selector 122a compares the first average black pixel section length with the first black pixel section length. In the first black pixel section, the ratio of the first black pixel section length Hh (i) to the first average black pixel section length Hh is smaller than the sixth reference value Q (for example, 0.5).
A black pixel section is selected as a first candidate section (S in FIG. 21).
3).

Here, the height and average value of each horizontal divided row are sequentially compared. If the height of a certain horizontal divided row is lower than a certain ratio with respect to the average, for example, the height H
When the value of 1 / Q times (for example, twice) of h (i) is lower than the average value Hh, it is highly likely that the horizontal divided row is erroneously divided. Therefore, the horizontal divided row is selected as a first candidate section as a candidate for integration correction.

In this embodiment, the divided row 150a shown in FIG. 22A is selected as the first candidate section.

When the height of each horizontal divided row is sequentially compared with the average value, and if the height does not satisfy the condition, the process sequentially proceeds to step S4 in FIG. In step S4, it is determined whether or not the parameter i is smaller than Nh. That is, it is determined whether or not only the height and the average value are compared for all the divided rows.

If the parameter i is not less than Nh, that is, if it is determined that the height and the average value have been compared for all the horizontal divided rows, the integration processing of the horizontal divided rows is ended.

If the parameter i is less than Nh, that is, if it is determined that there is still a horizontal division row whose height has not been compared with the average value, FIG.
To S5.

In the process of S5, the parameter i
Is incremented by one (increment). Then, the process proceeds to S3 in FIG. 21, and the next parameter i +
The height and the average value of the horizontal divided row corresponding to 1 are compared.

By the way, divided rows exist on both sides of the horizontal divided row 150a extracted as the first candidate section. For this reason, it is necessary to determine which of the two divided rows on both sides should be modified by integrating with the first candidate section. For how to make this decision,
This will be described in steps S6 to S10 described later.

Here, in order to facilitate understanding of the integration correction processing for the divided row 150a selected as the first candidate section in the step S3 in FIG. 21, FIG.
(A). FIG. 23A shows a horizontal division row 150.
It is a figure offered to explanation of unification correction processing of a.

Next, in the first integrated section length detecting section 124a, first, the horizontal division row 150a which is the first candidate section
The first nearest black pixel section on one side is referred to as a first candidate black pixel section 154. Here, in the horizontal divided row 150a,
The nearest first black pixel section above the plane of the paper is defined as a first candidate black pixel section 154. Then, the first candidate black pixel section 154
The black pixel section when the horizontal division row 150a and the horizontal divided row 150a are one continuous black pixel section is defined as a first integrated section. Then, the length Ht of the first integrated section 156 is obtained (FIG. 2).
1 S6).

As shown in FIG. 23A, the length Ht of the first integrated section 156 is the length of the horizontal divided row 150a, which is the first candidate section, and the length of the first candidate black pixel section 154. The total length is obtained by adding the length St of the first white pixel section 158 between the horizontal divided row 150a and the first candidate black pixel section 154.

Next, the first integrated section length error calculating section 125a
, The magnitude Et of the difference between the length Ht of the first integrated section 156 and the first average black pixel section length Hh is determined as the first integrated section length error (S7 in FIG. 21).

The first integrated section length error Et is given by the following equation (11).

Et = | Ht−Hh | (11) In the second integrated section length detecting unit 124b, first,
The nearest first black pixel section on the other side of the horizontal division row 150a that is the first candidate section is defined as a second candidate black pixel section. Here, the nearest first row below the horizontal plane of the horizontal divided row 150a is shown.
The horizontal divided row 150b that is a black pixel section is defined as a second candidate black pixel section 150b. When the second candidate black pixel section 150b and the horizontal divided row 150a are defined as one continuous black pixel section, the black pixel section is referred to as a second integrated section 16
Set to 0. Then, the length Hb of the second integrated section 160
(S8 in FIG. 21).

As shown in FIG. 23A, the length Hb of the second integrated section 160 is equal to the length of the horizontal divided row 150a, which is the first candidate section, and the length of the second candidate black pixel section 150b. The total length is the sum of the length Sb of the second white pixel section 162 between the horizontal divided row 150a and the second candidate black pixel section 150b.

Next, the second integrated section length error calculating section 125b
, The magnitude Eb of the difference between the length Hb of the second integrated section 160 and the first average black pixel section length Hh is determined as a second integrated section length error (S9 in FIG. 21).

The second integrated section length error Eb is given by the following equation (12).

Eb = | Hb−Hh | (12) Next, the first selected integrated section selection unit 126a
The integrated section length error Et and the second integrated section length error Eb are compared. Then, the integrated section having the smaller error is selected as the first selected integrated section (S10 in FIG. 21).

In the embodiment, when the first integrated section length error Et is equal to the second integrated section length error Eb, the second integrated section length error Eb is selected as the first selected integrated section. .

In this embodiment, the second integrated section length error Eb is shorter than the first integrated section length error Et. That is, Eb <Et. Therefore, in this embodiment, the second integrated section length error Eb is selected as the first selected integrated section (corresponding to the case of “Y” in S10 of FIG. 21).

Next, the second selected integrated section
In the integrated section 160, a second white pixel section 162 between the first candidate section 150a and a first black pixel section immediately adjacent to the first candidate section 150a (in this case, a second candidate black pixel section 150b). Is obtained (S11 in FIG. 21).

When one line is erroneously divided into two lines and cut out, the length of the white pixel section between the two divided lines is generally relatively short. On the other hand, the length of the white pixel section between appropriately cut out divided rows is relatively long. Therefore, when the length of the white pixel section is relatively long, there is a high possibility that the divided rows on both sides of the white pixel section are not erroneously divided rows. Therefore, in the next step S12, the length of the white pixel section is compared with a reference value.

Next, the length S of the second white pixel section 162
b is compared with the seventh reference value R (S12 in FIG. 21).

In this embodiment, since Sb <R, the flow advances to the next step S13 (corresponding to the case of “Y” in S12 in FIG. 21). When Sb ≧ R,
That is, in the case of “N” in S12 of FIG. 21, there is a high possibility that the first selected integrated section is not originally one line. For this reason, the process proceeds to step S4 without correcting the line segmentation for the first candidate section.

Next, the length S of the second white pixel section 162 is calculated.
When b is shorter than the seventh reference value R, the second integrated section 160 as the first selected integrated section is set as one continuous first integrated section.
The line segmentation is corrected by setting the black pixel section (S13 in FIG. 21).

In this embodiment, two horizontal divided rows 1
The correction is performed by cutting out the second integrated section 160 including 50a and 150b as one horizontal divided row 150. This horizontal division line 150 includes three characters of "two", "kichi" and "go".

Although not applicable in this embodiment, if Et <Eb in S10, the first
The first integrated section 156 is selected as the selected integrated section. Then, in this case, for the first integrated section 156,
The process of steps S14 to S16 is performed.

In this case, first, the length St of the first white pixel section 158 is detected (S14 in FIG. 21).

Next, the length St of the first white pixel section 158 is compared with the seventh reference value R (S15 in FIG. 21). Note that S
If t ≧ R, the process proceeds to step S4.

Next, when the length St of the first white pixel section 158 is shorter than the seventh reference value R (corresponding to the case of “Y” in S15), the first selected integrated section as the first selected integrated section. The line segmentation is corrected by integrating one integrated section 156 as one continuous first black pixel section (S16 in FIG. 21).

(Regarding the Correction of Vertically Split Lines)
With reference to FIG. 24, an example of a method of integrating and correcting the vertically divided rows 152 shown in FIG. 22B will be described. FIG.
8 is a flowchart provided for describing a method of integrating and correcting vertical divided rows.

In general, the width of a vertically divided row that is erroneously divided into two and cut out is narrower than that of another appropriately cut out divided row, and is about half that of the other divided rows. further,
The line interval between two erroneously cut out divided lines is generally smaller than the line interval between the other divided lines. Therefore, if attention is paid to the width or the width and the line spacing of the cut-out divided lines, a divided line that is erroneously cut out as two vertical divided lines where the original cut-out line should be cut out as one vertical divided line is found. It is possible to correct the split line segmentation.

Therefore, in this embodiment, in performing the integrated correction of the vertically divided rows, first, in the second average black pixel section length calculating section 120b, the second divided row extracted by the second divided row extracting section 18b The arithmetic mean of the length (second black pixel section length) of the second black pixel section of the row (vertical divided row) is determined as the second average black pixel section length (S1 in FIG. 24).

When the second divided row is a vertical divided row, the second divided row
The second black pixel section length of the divided row indicates the width of the vertical divided row.
Then, in this case, the second average section length Wv represents an average value of the widths of the vertically divided rows.

Next, in order to sequentially check whether or not each vertically divided row is erroneously cut out, a parameter j for giving the serial number of the vertically divided row to the second candidate section selection unit 122b for convenience is given to the second candidate section selection unit 122b. prepare. Then, 1 is substituted for this parameter j (S2 in FIG. 24).

The maximum value of the parameter j is the total number Nv of the vertically divided rows.

Next, the second candidate section selector 122b compares the second average black pixel section length with the second black pixel section length. Then, in the second black pixel section, the ratio of the second black pixel section length Wv (j) to the second average black pixel section length Wv is smaller than the eighth reference value Q (for example, 0.5).
The black pixel section is selected as the second candidate section (S in FIG. 24).
3).

Here, the width of each vertical divided row and the average value are sequentially compared. When the width of a certain vertical divided row is smaller than a certain ratio with respect to the average value, for example, the width Wv
If the value of 1 / Q (for example, twice) of (j) is lower than the average value Wv, it is highly likely that the vertically divided row is erroneously divided. Therefore, the vertically divided row is selected as a second candidate section as a candidate for integration correction.

In this embodiment, the divided row 152a shown in FIG. 22B is selected as a second candidate section.

When sequentially comparing the width of each vertical divided row with the average value, if the width does not satisfy the condition, the flow sequentially advances to step S4 in FIG. In step S4, it is determined whether or not the parameter j is smaller than Nv. That is, it is determined whether or not only the width and the average value are compared for all the divided rows.

If the parameter j is not less than Nv, that is, if it is determined that the width and the average value have been compared for all the vertically divided rows, the integration processing of the vertically divided rows is ended.

If the parameter j is smaller than Nv, that is, if it is determined that there is still a vertically divided row whose width has not been compared with the average value, the process proceeds to S5 in FIG. .

In the process of S5, the parameter j
Is incremented by one (increment). Then, the process proceeds to S3 in FIG. 24, where the next parameter j +
The width and the average value of the vertical division row corresponding to 1 are compared.

By the way, divided rows exist on both sides of the vertical divided row 152a extracted as the second candidate section. Therefore, it is necessary to determine which of the two divided rows on both sides is to be modified by integrating with the second candidate section. For how to make this decision,
This will be described in steps S6 to S10 described later.

Here, in order to facilitate understanding of the integration correction processing for the divided row 152a selected as the second candidate section in the step S3 of FIG. 24, FIG.
(B). FIG. 23B shows a vertical division row 152.
FIG. 13B is a diagram provided for explanation of an integrated correction process of FIG.

Next, in the third integrated section length detecting section 124c, first, the vertical division row 152a as the second candidate section
The second black pixel section nearest to one side is referred to as a third candidate black pixel section 164. Here, the vertical division row 152a
The nearest second black pixel section on the left side of the paper is referred to as a third candidate black pixel section 164. Then, the third candidate black pixel section 164
The black pixel section when the vertical division row 152a and the vertical divided row 152a are one continuous black pixel section is referred to as a third integrated section 166. Then, the length Wl of the third integrated section 166 is determined (S6 in FIG. 24).

The length Wl of the third integrated section 166 is, as shown in FIG. 23B, the length of the vertical divided row 152a which is the second candidate section and the length of the third candidate black pixel section 164. The total length is the sum of the length S1 of the third white pixel section 168 between the vertical divided row 152a and the third candidate black pixel section 164.

Next, the second integrated section length error calculating section 125b
24, the magnitude El of the difference between the length Wl of the third integrated section 166 and the second average black pixel section length Wv is determined as the third integrated section length error (S7 in FIG. 24).

The third integrated section length error El is given by the following equation (13).

El = | Wl−Wv | (13) In the fourth integrated section length detecting section 124d, first,
The nearest second black pixel section on the other side of the vertical division row 152a that is the second candidate section is defined as a second candidate black pixel section. Here, the second divided line 152a on the right side of the drawing in the vertical divided row 152a
The vertical divided row 152b that is a black pixel section is defined as a second candidate black pixel section 152b. Then, when the second candidate black pixel section 152b and the vertical divided row 152a are defined as one continuous black pixel section, the black pixel section is referred to as a fourth integrated section 17.
Set to 0. The length Wr of the fourth integrated section 170
(S8 in FIG. 24).

The length Wr of the fourth integrated section 170 is, as shown in FIG. 23B, the length of the vertical divided row 152a which is the second candidate section and the length of the fourth candidate black pixel section 152b. The total length is obtained by adding the length Sr of the fourth white pixel section 172 between the vertical divided row 152a and the fourth candidate black pixel section 152b.

Next, the second integrated section length error calculating section 125b
, The magnitude Er of the difference between the length Wr of the fourth integrated section 170 and the second average black pixel section length Wv is determined as the fourth integrated section length error (S9 in FIG. 24).

The fourth integrated section length error Er is given by the following equation (14).

Er = | Wr−Wv | (14) Next, the second integrated section length error El is compared with the fourth integrated section length error Er in the second integrated section length selecting section 126b. Then, the integrated section having the smaller error is selected as the second selected integrated section (S10 in FIG. 24).

In the embodiment, when the third integrated section length error El is equal to the fourth integrated section length error Er, the fourth integrated section length error Er is selected as the second selected integrated section. .

[0408] In this embodiment, the fourth integrated section length error Er is shorter than the third integrated section length error El. That is, Er <El. Therefore, in this embodiment, the third integrated section length error Er is selected as the second selected integrated section (corresponding to the case of “Y” in S10 of FIG. 24).

[0409] Next, the fourth selected integrated section is referred to as the fourth selected integrated section.
The fourth white pixel section 172 between the second candidate section 152a and the second black pixel section immediately adjacent to the second candidate section 152a (in this case, the second candidate black pixel section 152b) in the integrated section 170. Is determined (S11 in FIG. 24).

When one line is erroneously divided into two lines and cut out, the length of the white pixel section between the two divided lines is generally relatively short. On the other hand, the length of the white pixel section between appropriately cut out divided rows is relatively long. Therefore, when the length of the white pixel section is relatively long, there is a high possibility that the divided rows on both sides of the white pixel section are not erroneously divided rows. Therefore, in the next step S12, the length of the white pixel section is compared with a reference value.

Next, the length S of the fourth white pixel section 172
r is compared with the ninth reference value R (S12 in FIG. 24).

In this embodiment, since Sr <R, the process proceeds to the next step S13 (corresponding to the case of “Y” in S12 in FIG. 24). Also, when Sr ≧ R,
That is, in the case of “N” in S12 of FIG. 24, there is a high possibility that the second selected integrated section is not originally one line. For this reason, the process proceeds to step S4 without correcting the line segmentation for the second candidate section.

Next, the length S of the fourth white pixel section 172 is calculated.
When r is shorter than the ninth reference value R, the fourth integrated section 170 as the second selected integrated section is set as one continuous second integrated section.
Correction of line segmentation is performed by setting a black pixel section (S13 in FIG. 24).

In this embodiment, two vertically divided rows 1
The fourth integrated section 170 including 52a and 152a is corrected by cutting it out as one vertical divided row 152. This vertical division line 152 includes three characters of “eight”, “north”, and “thanks”.

[0415] Although this is not applicable in this embodiment, if El <Er in S10, the second
The third integrated section 166 is selected as the selected integrated section. Then, in this case, for the third integrated section 166,
The process of steps S14 to S16 is performed.

In this case, first, the length S1 of the third white pixel section 168 is detected (S14 in FIG. 24).

Next, the length S1 of the third white pixel section 168 is compared with the ninth reference value R (S15 in FIG. 24). Note that S
If 1 ≧ R, the process proceeds to step S4.

Next, when the length S1 of the third white pixel section 168 is shorter than the ninth reference value R (corresponding to the case of “Y” in S15), the second selection integration section is used. The line segmentation is corrected by integrating the three integrated sections 166 as one continuous second black pixel section (S16 in FIG. 24).

(Operation Example of Division Correction) Next, an example of a correction method in the case where one divided row is cut out as two divided rows by mistake will be described.

FIGS. 26A and 26B show examples of horizontally divided horizontal lines to be corrected. FIG. 26A shows a horizontal divided line 180 in which the upper and lower divided lines are erroneously cut out as one divided line because the line interval is narrow and a part of the two divided lines overlaps. . In the upper part of the horizontal division line 180, "symbo"
l ”, and“ calcu ”
rate ". And this "s
ymbol ", the second character" y "of the character string, and" cal "
The third character “l” of the character string “culate” partially overlaps with each other. And this "symbol"
And the character string "calculate" are to be cut out as individual divided lines.

In FIG. 26B, since the row direction of the document image is oblique to the scanning direction, each black pixel section of the upper and lower two rows in the projection distribution is continuous. The horizontal divided row 182 in which the upper and lower divided rows are erroneously cut out as one divided row is shown. The upper part of the horizontal divided line 182 contains the character string “symbol”, and the lower part thereof contains the character string of “calculate”. Each of these character strings is inclined to the upper right with respect to the divided row 182 whose outline is indicated by a rectangular frame.

In this case, the first direction is the horizontal direction (horizontal direction on the paper, the direction along the X axis), and the second direction is the vertical direction (vertical direction on the paper, the direction along the Y axis). Therefore,
The process of dividing and correcting the horizontally divided horizontal row 180 shown in FIG. 26A is performed by the first dividing and correcting unit 92a. If the second direction is the horizontal direction, the process of correcting the division of the horizontal divided row 180 is performed by the second
b.

(Partial Correction of Horizontal Divided Lines 1) First,
With reference to FIG. 25, an example of the division correction method of the horizontal division row 180 shown in FIG. 26A will be described. FIG.
9 is a flowchart for explaining a division correction method of a horizontally divided row.

In general, when two horizontal divided lines are erroneously cut out as one horizontal divided line, the height of the erroneously cut out horizontal divided line is larger than that of another appropriately divided horizontal divided line. And is about twice as high as other divided rows.
Therefore, if attention is paid to the height of the cut-out divided lines, it is possible to find a divided line that is erroneously cut out as one divided line where it should originally be cut out as two divided lines, and correct the divided line cut-out. Becomes

Therefore, in this embodiment, in performing the division correction of the horizontal division line, first, in the first average black pixel section length calculation unit 131a, the first division line extracted by the first division line extraction unit 18a The arithmetic mean of the length of the first black pixel section (first black pixel section length) of the row is determined as the first average black pixel section length (S1 in FIG. 25). When the first divided row is a horizontal divided row, the first black pixel section length of the first divided row indicates the height of the horizontal divided row. And in this case,
The first average section length Hh represents an average value of the heights of the horizontal divided rows.

Next, in order to sequentially check whether or not each horizontally divided row is erroneously cut out, the first division candidate section selecting unit 132a determines the serial number of the horizontally divided row for convenience. A parameter i to be given is prepared. Then, 1 is substituted for this parameter i (FIG. 2).
5 S2).

The maximum value of this parameter i is the total number Nh of horizontal divided rows.

Next, the first division candidate section selector 132a compares the first average black pixel section length with the first black pixel section length. Then, of the first black pixel sections, the first black pixel section length Hh with respect to the first average black pixel section length Hh
The first black pixel section in which the ratio of (i) is smaller than the eleventh reference value U (for example, 2) is selected as the first candidate section (FIG. 2).
5 S3).

Here, the height of each horizontal divided row and the average value are sequentially compared. If the height of a certain horizontal divided row is lower than a certain ratio with respect to the average, for example, the height H
The value of 1 / Q times (for example, 1/2 times) of h (i) is the average value H
If h is higher than h, it is highly likely that the two divided lines were originally incorrectly divided into one divided line.
For this reason, the horizontal divided row is selected as a first candidate section as a candidate for division correction.

Then, in this embodiment, the divided row 180 shown in FIG. 26A is selected as the first candidate section.

When the height and average value of each horizontal divided row are sequentially compared, and if the height does not satisfy the condition, the process proceeds to step S4 in FIG. 25 sequentially. In step S4, it is determined whether or not the parameter i is smaller than Nh. That is, it is determined whether the height and the average value have been compared for all the divided rows.

If the parameter i is not less than Nh, that is, if it is determined that the height and the average value have been compared for all the horizontal divided rows, the process of integrating the horizontal divided rows is ended.

If the parameter i is less than Nh, that is, if it is determined that there is still a horizontal division row whose height has not been compared with the average value, FIG.
To S5.

In the process of S5, the parameter i
Is incremented by one (increment). Then, the process proceeds to S3 in FIG. 25, where the next parameter i +
The height and the average value of the horizontal divided row corresponding to 1 are compared.

Next, the following steps S6 to S8 are performed in order to determine a scanning range for obtaining a first position and a second position described later. The state of this processing is shown in FIGS.
It is shown in (C).

First, a preliminary upper limit Yhs is set at a position lower than the upper end of the horizontal divided row 180 by a distance of Hh / 2 (half the average black pixel section length) which is half the average height. This preliminary upper limit Yhs is given by the Y coordinate. Further, a preliminary lower limit Yhe is set at a position higher than the lower end of the horizontal divided row 180 by a distance of Hh / 2 which is half of the average height. This preliminary lower limit Yhe is given by the Y coordinate. (S6 in FIG. 25) ((A) in FIG. 27).

Next, a portion of the horizontal divided row 180 between the preliminary upper limit Yhs and the preliminary lower limit Yhe is scanned by using the X-axis direction as the scanning direction and the Y-axis direction as the sub-scanning direction. The partial projection distribution 178 is created by detecting the cumulative number of black pixels for each line. Then, in the partial projection distribution 178, the minimum position Yd at which the cumulative number of black pixels is minimum is obtained.
This minimum position Yd is given by the Y coordinate (S in FIG. 25).
7) ((B) of FIG. 27).

Next, an upper limit Yds is set at a position higher than the minimum position Yd of the horizontal divided row 180 by a distance of Hh / 2 which is half of the average height. This upper limit Yds is given by the Y coordinate. Also, the minimum position Y of the horizontal divided row 180
A lower limit Yde is set at a position lower than d by a distance of Hh / 2 which is half of the average height. This lower limit Yde is
It is given by the Y coordinate (S8 in FIG. 25) ((C) in FIG. 27).

In this manner, the upper limit Yds and the lower limit Yde of the scanning range for determining the first and second positions are determined. Hereinafter, with reference to FIGS. 28A and 28B, a process of detecting the first and second positions in the scanning range will be described.

Next, in the first position extracting section 134a,
Among the positions where the black pixels are detected on the scanning line that scans the vicinity of the center of the first division candidate section among the scanning lines, the position closest to one side along the first direction is determined as the first position. Then, in the nearest first position extracting unit 136a, the first position closest to the other side in the first direction among the first positions is determined as the nearest first position (S9 in FIG. 25).

In this embodiment, the left side of the sheet of FIG. 28 is referred to as “one-sided”, and the right side of the sheet of FIG. 28 is referred to as “the other side”. Therefore, the first position is a position where the scanning range between the upper limit Yds and the lower limit Yde is scanned from left to right, and a black pixel is first detected for each scanning line. Further, the nearest first position is the rightmost position among the first positions.

In this embodiment, the nearest first position is detected, and its coordinates (Xr, Yr) are obtained (FIG. 28).
(A)).

Next, in the second position extracting section 134b,
Of the positions where black pixels are detected on the scanning line that scans the vicinity of the center of the first division candidate section among the scanning lines, the position closest to the other side in the second direction is determined as the second position. Then, the nearest second position extracting unit 136b obtains, as the nearest second position, the second position closest to the first position along the second direction among the second positions (S10 in FIG. 25).

In this embodiment, the second position is
The scanning range between the upper limit Yds and the lower limit Yde is scanned from right to left, and a position where a black pixel is first detected is set for each scanning line. The second closest position is the leftmost position among the second positions.

In this embodiment, the nearest second position is detected, and its coordinates (X1, Y1) are obtained (FIG. 28).
(A)).

Next, the first nearest position and the second nearest position are compared (S11 in FIG. 25). The X coordinate (Xr) of the nearest first position (Xr, Yr) is one side (left side) of the X coordinate (Xl) of the nearest second position (Xl, Yl).
(That is, Xr <Xl), or the x coordinate (Xl) of the nearest first position and the X coordinate (Xr) of the nearest second position are located on the same straight line along the second direction. Is determined (ie, Xr = X1). That is, it is determined whether or not Xr ≦ X1 is satisfied.

If Xr ≦ X1, the first division section 138a divides the first division candidate section into two at the center of the first division candidate section to generate two divided black pixel sections. Thus, the line cutout is corrected (S12 in FIG. 25).

In the horizontal division line 180, Xr <Xl
It is. Therefore, here, the horizontal divided row 180 is forcibly divided vertically into two by a dividing line along the minimum position Yd ((B) in FIG. 28). This division generates two horizontal divided rows 180a and 180b. The upper horizontal divided line 180a includes the character string “symbol”. The lower horizontal divided row 180b includes “c
Calculate ”is included.

In this manner, as in the case of the horizontal divided row shown in FIG. 26A, the cutout of a divided row that is erroneously integrated and cut out because the divided rows are close to each other is corrected. In many cases, the line segmentation can be corrected by forcibly dividing the horizontal divided line into two in the vertical direction.

(Partial Correction of Horizontal Divided Lines Part 2)
With reference to FIG. 25, an example of the division correction method of the horizontal division row 182 shown in FIG. 26B will be described. In this case, the processing contents up to S11 in FIG.
(A) is the same as the processing content in the case of the division correction, and a detailed description thereof will be omitted.

Here, after a horizontal division row 182 as a division candidate black pixel section is obtained, a first position and a second position
In order to determine the scanning range when obtaining the position, the following S6 to
The processing of S8 is performed. The state of this processing is shown in FIGS.

First, a preliminary upper limit Yhs is set at a position lower than the upper end of the horizontal divided row 182 by a distance of Hh / 2 (half the average black pixel section length) which is half the average height. This preliminary upper limit Yhs is given by the Y coordinate. Further, a preliminary lower limit Yhe is set at a position higher than the lower end of the horizontal divided row 180 by a distance of Hh / 2 which is half of the average height. This preliminary lower limit Yhe is given by the Y coordinate. (S6 in FIG. 25) ((A) in FIG. 29).

Next, in the horizontal divided row 182, a portion in the range between the preliminary upper limit Yhs and the preliminary lower limit Yhe is scanned by using the X-axis direction as the scanning direction and the Y-axis direction as the sub-scanning direction. A partial projection distribution 176 is created by detecting the cumulative number of black pixels for each line. Then, in the partial projection distribution 176, the minimum position Yd at which the cumulative number of black pixels is minimum is obtained. This minimum position Yd is given by the Y coordinate (FIG. 25)
S7) (FIG. 29B).

Next, from the minimum position Yd of the horizontal divided row 180. An upper limit Yds is set at a position raised by a distance of Hh / 2 which is half of the average height. This upper limit Yds is given by the Y coordinate. Also, the minimum position Y of the horizontal divided row 180
A lower limit Yde is set at a position lower than d by a distance of Hh / 2 which is half of the average height. This lower limit Yde is
It is given by the Y coordinate (S8 in FIG. 25) ((C) in FIG. 29).

[0455] In this manner, the upper limit Yds and the lower limit Yde of the scanning range for determining the first and second positions are determined. Hereinafter, with reference to FIGS. 30A and 30B, a process of detecting the first and second positions within the scanning range will be described.

Next, in the first position extracting section 134a,
Among the positions where the black pixels are detected on the scanning line that scans the vicinity of the center of the first division candidate section among the scanning lines, the position closest to one side along the first direction is determined as the first position. Then, in the nearest first position extracting unit 136a, the first position closest to the other side in the first direction among the first positions is determined as the nearest first position (S9 in FIG. 25).

In this embodiment, the left side of FIG. 30 is referred to as “one-sided”, and the right side of FIG. 30 is referred to as “other-side”. Therefore, the first position is a position where the scanning range between the upper limit Yds and the lower limit Yde is scanned from left to right, and a black pixel is first detected for each scanning line. Further, the nearest first position is the rightmost position among the first positions.

In this embodiment, the nearest first position is detected, and its coordinates (Xr, Yr) are obtained (FIG. 30).
(A)).

Next, in the second position extracting section 134b,
Of the positions where black pixels are detected on the scanning line that scans the vicinity of the center of the first division candidate section among the scanning lines, the position closest to the other side in the second direction is determined as the second position. Then, the nearest second position extracting unit 136b obtains, as the nearest second position, the second position closest to the first position along the second direction among the second positions (S10 in FIG. 25).

[0460] In this embodiment, the second position is:
The scanning range between the upper limit Yds and the lower limit Yde is scanned from left to right, and for each scanning line, a position where a black pixel is detected first is set. The second closest position is the leftmost position among the second positions.

In this embodiment, the nearest second position is detected, and its coordinates (X1, Y1) are obtained (FIG. 30).
(A)).

Next, the first nearest position and the second nearest position are compared. (S11 in FIG. 25). The X coordinate (Xr) of the nearest first position (Xr, Yr) is on the other side (right side) of the X coordinate (Xl) of the nearest second position (Xl, Yl).
Is determined (ie, Xl <Xr). Here, as shown in FIG. 30A, Xl <X
r is satisfied.

If Xl <Xr, the second dividing section 138b makes a first division by a dividing line passing through the middle between the nearest first position and the nearest second position and extending in the second direction. The horizontal division row 182 as a candidate section is divided. Here, the horizontal division row 182 is divided into two at the position where the X coordinate is (Xr + X1) / 2.

By this division, a first divided region 184 on one side (left side) and another divided region 186 on the other side (right side) are generated.

Then, the one divided area 184 is further divided by the scanning line from which the nearest second position (Xr, Yr) is obtained, and divided into two divided black pixel sections 184a and 184b.
Generate That is, the one divided region 184 is vertically divided into two at the position of Yr. By this division, an upper left divided row 184a and a lower left divided row 184b are generated.

The other divided area 186 is further divided by the scanning line from which the nearest first position (X1, Y1) has been obtained to generate two divided black pixel sections 186a and 186b. That is, the other divided area 186 is vertically divided into two at the position of Yl. By this division, the upper right division line 186
a and the lower right split row 186b are generated.

The line segmentation is corrected by generating a total of four divided black pixel sections in this manner (FIG. 25).
S13) ((B) of FIG. 30).

As described above, as in the case of the horizontal division line shown in FIG. 26B, the division direction erroneously integrated and cut out because the line direction of the document image is inclined with respect to the scanning direction. In many cases where you want to fix line breaks,
The line segmentation can be corrected by dividing the divided line into four.

The upper left divided row 184a and the upper right divided row 18
6a are connected to each other in the process of connecting the divided rows. Further, the lower left divided row 184b and the lower right divided row 186b are connected to each other at the time of the processing of linking the divided rows.

(Example of operation of connection of divided rows)
An operation example of the connection of the divided rows will be described.

(Regarding Operation of Connecting Horizontally Divided Rows) First, an example of connection between horizontally divided rows will be described with reference to FIGS. 31 (A) and 31 (B). Here, the horizontal direction is a first direction, and the vertical direction is a second direction. Also, in FIGS. 31A and 31B, X
The axis is taken and the Y axis is taken in the vertical direction.

[0472] Fig. 31A shows two horizontal divided rows 190a and 190b included in divided areas adjacent to each other, respectively. In the horizontal divided rows 190a and 190b, portions of the first black pixel section overlap each other. That is, the horizontal division rows 190a and 19
0b are adjacent to each other with a slight shift.

In the first divided row connection section 26a, when the length Kh of the first black pixel section overlapping the horizontal divided rows 190a and 190b is longer than the fourth reference value V, the horizontal Connect the split lines. The value of the fourth reference value V is, for example, It is preferable that the height be about half of the average height of the horizontal divided rows. Here, since Kh> V is satisfied, the horizontal divided rows 190a and 190b are connected to each other.

With this connection, a row 190 is generated as shown in FIG. In FIG. 31B, for the sake of convenience, the outline of the row 190 is divided into the horizontal divided row 1
It is shown slightly larger than the contours of 90a and 190b. However, in practice, the contour of row 190 matches the contour of horizontal split rows 190a and 190b.

(Regarding Operation of Connecting Vertically Divided Rows) Next, an example of connection between vertically divided rows will be described with reference to FIGS. 32 (A) and 32 (B). Here, the horizontal direction is a second direction, and the vertical direction is a first direction. Also, in FIGS. 32A and 32B, X is set in the horizontal direction.
The axis is taken and the Y axis is taken in the vertical direction.

FIG. 32A shows two vertical divided rows 192a and 192b included in divided areas adjacent to each other. In the vertical divided rows 192a and 192b, parts of the second black pixel section overlap each other. That is, the vertical divided rows 192a and 19
2b are adjacent to each other with a slight shift.

In the second divided row connection section 26b, when the length Kw of the second black pixel section overlapping the vertical divided rows 192a and 192b is longer than the fifth reference value V, Connect the split lines. The value of the fifth reference value V is, for example, It is preferable to set the height to about half of the average height of the vertical divided rows. Here, since Kw> V is satisfied, the vertically divided rows 192a and 192b are connected to each other.

With this connection, a row 192 is generated as shown in FIG. In FIG. 32B, for the sake of convenience, the outline of the row 192 is divided into the vertical division row 1
It is shown slightly larger than the contours 92a and 192b. However, in practice, the contour of row 192 matches the contour of vertical split rows 192a and 192b.

In each of the embodiments described above, only examples in which the present invention is configured under specific conditions have been described. However, the present invention can be subjected to many changes and modifications. For example, in the embodiment described above, the horizontal direction is the first direction,
Although the vertical direction has been described as the second direction, in the present invention, for example, the vertical direction may be the second direction, and the horizontal direction may be the first direction. Further, the first direction and the second direction need only be orthogonal to each other, and need not necessarily be along the vertical direction and the horizontal direction of the document area of the document image.

In the above-described embodiment, the first and second reference values for obtaining the black pixel section in the projection distribution are each one pixel. However, in the present invention, the first and second reference values are used as the first and second reference values. Any suitable natural number can be taken. For example, several pixels may be used as the reference value. Further, the first reference value and the second reference value may be different from each other.

In the above-described embodiment, the same number of the first direction divided areas and the second direction divided areas are generated. However, in the present invention, the number of the first direction divided areas and the number of the second direction divided areas are different. Are not necessarily the same.

In the above-described embodiment, the first correction unit 24a and the second correction unit 24b are provided. However, the line segmenting apparatus and method of the present invention can be used even when these correction units are not provided. Can be applied.

Although the line direction confirmation switching unit 12 is provided in the above-described embodiment, the line segmenting device of the present invention can be applied to a case where the line direction confirmation switching unit 12 is not provided. In that case, for example, the operator
The transmission destination of the document image information may be switched by checking the information in the row direction.

Although the switching output unit 22 is provided in the above-described embodiment, the line segmenting device according to the present invention
The present invention can be applied to a case where the switching output unit 22 is not provided. In this case, for example, the operator may switch the transmission destination of the document image after confirming the result of the line direction determination.

[0485]

【The invention's effect】

(Line cutting method and line cutting device) According to the line cutting method and line cutting device of the present invention, when the line direction of a document image is unknown, the first direction obtained in a process as a part of the line cutting process. Using the information of the first divided row cut out based on the projection distribution and the information of the second divided row cut out based on the second direction projection distribution obtained in the process of part of the row cutout processing, Determine the row direction. After the row direction is determined, the divided rows corresponding to the determined row direction are connected to each other to cut out the row. Therefore, according to the present invention, even when the line direction of a document image is unknown, the information obtained in the line cutout process can be obtained without performing the process of determining the line direction separately from the line cutout in advance. The row direction can be determined by using this to perform line segmentation.

(First Line Cutout Correction Method and First Line Cutout Correction Device) In addition, according to the first line cutout correction method and the first line cutout correction device of the present invention, one line is incorrectly set. Correction of line cutout when cut out as two lines can be performed.

(Second line cutout correcting method and second line cutout correcting device) According to the second line cutout correcting method and the second line cutout correcting device of the present invention, two individual lines are It is possible to correct the line cutout when it is cut out as one line by mistake.

[Brief description of the drawings]

FIG. 1 is a functional block diagram for explaining a configuration of a line segmentation device according to an embodiment;

FIG. 2 is a functional block diagram for explaining the configuration of a character recognition device.

FIG. 3 is a functional block diagram for explaining a configuration of a row direction determination unit.

FIG. 4 is a functional block diagram for explaining a configuration of a first determination unit;

FIG. 5 is a functional block diagram for explaining a configuration of a second determination unit;

FIG. 6 is a functional block diagram for explaining a configuration of a third determination unit;

FIG. 7 is a functional block diagram for explaining a configuration of a first correction unit;

FIG. 8 is a functional block diagram for explaining a configuration of a first integrated correction unit;

FIG. 9 is a functional block diagram for explaining the configuration of a first division correction unit.

FIG. 10 is a functional block diagram for explaining a configuration of a second correction unit;

FIG. 11 is a functional block diagram for explaining a configuration of a second integrated correction unit;

FIG. 12 is a functional block diagram for explaining a configuration of a second division correction unit.

FIG. 13 is a flowchart for explaining the operation of the line cutout device;

FIG. 14 is a document image to be determined.

FIGS. 15A and 15B are diagrams for explaining a divided area; FIGS.

FIGS. 16A and 16B are diagrams for explaining a divided region and its projection distribution.

FIG. 17 is a flowchart for explaining the operation of a determination unit;

FIGS. 18A to 18D are views for explaining the operation of the first determination unit (first direction).

FIGS. 19A to 19C are diagrams for explaining the operation of the first determination unit (second direction).

FIGS. 20A and 20B are diagrams for explaining the calculation of the variation of the black pixel section length and the white pixel section length of the divided area.

FIG. 21 is a flowchart for explaining the operation of integrating and correcting horizontal divided rows.

FIGS. 22A and 22B are examples of a target of integrated correction;

FIGS. 23A and 23B are diagrams for explaining an integrated correction method;

FIG. 24 is a flowchart for explaining the operation of integrating and correcting vertical divided rows.

FIG. 25 is a flowchart for explaining the operation of dividing and correcting a horizontally divided row.

FIGS. 26A and 26B are examples of target of division correction.

FIGS. 27A to 27C are diagrams (part 1) for describing a first modification example of division;

FIGS. 28A and 28B are diagrams (part 2) for explaining the first modification example following FIG. 27C;

FIGS. 29A to 29C are diagrams (part 1) for describing a second modification example;

FIGS. 30A and 30B are diagrams (part 2) for explaining the second modification example following FIG. 29C;

FIGS. 31A and 31B are diagrams for explaining connection of horizontal divided rows.

FIGS. 32A and 32B are diagrams for explaining connection of a vertically divided row.

[Explanation of symbols]

10: Line segmentation device 12: Line direction confirmation switching unit 12a: Line direction confirmation unit 12b: Switching unit 14a: First direction division region generation unit 14b: Second direction division region generation unit 16a: First direction projection distribution generation unit 16b : Second direction projection distribution creating unit 18a: first divided line segmenting unit 18b: second divided line segmenting unit 20: line direction determining unit 22: switching output unit 24a: first modifying unit 24b: second modifying unit 26a: second 1-divided line connection unit 26b: second divided-line connection unit 28: control unit 30: scanner 31: image memory 32: character area setting unit 33: character cutout unit 34: character recognition unit 36a: first division region number counting unit 36b : Second black area section counting section 38a: first black pixel section number counting section 38b: second black pixel section number counting section 40: section number division number comparison determination section 42a: first variation calculation section 42b: second frame Fluctuation calculation unit 44: Variation comparison determination unit 46a: First average value calculation unit 46b: Second average value calculation unit 48: Average value comparison determination unit 50: Text area 50a: First division area 50b: Second division area 50c : Third divided area 50d: Fourth divided area 50e: Fifth divided area 50f: Sixth divided area 52: Character string, line 52a: First divided line 52b: Second divided line 52c: Third divided line 54a, 54b : First direction projection distribution 54e: second direction projection distribution 56a, 56b: black pixel section 60: document image 62: character string, line 64a: full-width character 64b: half-width character 66: character area 68: first direction division area 68a : First divided area 68b: Second divided area 68c: Third divided area 70: Second divided area 70a: Fourth divided area 70b: Fifth divided area 70c: Sixth divided area 70d: Seventh Split area 72: Projection distribution in the first direction 74: Projection block 74a: First projection block 74b: Second projection block 74c: Third projection block 74j: Tenth projection block 76: Split row, horizontal split row 78: Second direction Projection distribution 78a: First projection 78b: Second projection 78c: Third projection 78d: Fourth projection 78e: Fifth projection 80: Split row, vertical split row 82: First black pixel section 84: Second black pixel section 86 : First white pixel section 88: second white pixel section 90a: first integrated correction section 90b: second integrated correction section 92a: first division correction section 92b: second division correction section 120a: first average black pixel section length Calculator 120b: Second average black pixel section length calculator 122a: First candidate section selector 122b: Second candidate section selector 124a: First integrated section length detector 124b: Second integrated section length detector Section 124c: third integrated section length detecting section 124d: fourth integrated section length detecting section 125a: first integrated section length error calculating section 125b: second integrated section length error calculating section 125c: third integrated section length error calculating section 125d : 4th integrated section length error calculating section 126a: 1st selected integrated section length selecting section 126b: 2nd selected integrated section length selecting section 128a: 1st white pixel section length detecting section 128b: 2nd white pixel section length detecting section 130a : 1st integration section 130b: 2nd integration section 131a: 1st average black pixel section length calculation section 131b: 2nd average black pixel section length calculation section 132a: 1st division candidate section selection section 132b: 2nd division candidate section selection Unit 134a: first position extraction unit 134b: second position extraction unit 134c: third position extraction unit 134d: fourth position extraction unit 136a: nearest first position extraction unit 136b: second nearest position Extraction unit 136c: Nearest third position extraction unit 136d: Nearest fourth position extraction unit 138a: First division unit 138b: Second division unit 138c: Third division unit 138d: Fourth division unit 150, 150a, 150b: Division row 152, 152a, 152b: Division row 154: First candidate black pixel section 156: First integration section 158: First white pixel section 160: Second integration section 162: Second white pixel section 164: Third candidate black Pixel section 166: Third integrated section 168: Third white pixel section 170: Fourth integrated section 172: Fourth white pixel section 176, 178: Partial projection distribution 180, 180a, 180b, 182, 182a, 18
2b: horizontal divided line 184: one-side divided region 184a: upper left divided line 184b: lower left divided line 186: other divided region 186a: upper right divided line 186b: lower right divided line 190: horizontal line 190a, 190b: horizontal divided line 192: vertical Rows 192a and 192b: Vertically divided rows 201: First determination unit 202: Second determination unit 203: Third determination unit

Claims (27)

[Claims] (A1) dividing the document image by a dividing line along a second direction when the line direction of the document image is unknown and when the line direction of the document image is the first direction. A process of generating a plurality of first direction divided regions arranged along a first direction orthogonal to the second direction, and (b1) when the first direction divided regions are generated, the first direction Each of the divided areas is scanned in the first direction, the number of cumulative black pixels for each scan is detected, and the distribution of the cumulative number of black pixels along the second direction is determined to project in the first direction. And (c1) when the first direction projection distribution is created, the cumulative number of black pixels equal to or greater than a first reference value in each of the first direction projection distributions along the second direction. Each continuous section is defined as a first black pixel section, A process in which a scanning area where the first black pixel section is obtained is used as a first divided line to perform divided line segmentation; (a2) When the line direction of the document image is unknown, and when the line direction of the document image is the second direction A process of generating a plurality of second-direction divided regions arranged along the second direction by dividing the document image by dividing lines along the first direction; and When a two-way divided area is generated, each of the second-directional divided areas is scanned along the second direction to detect the cumulative number of black pixels for each scan, and A process of creating a second direction projection distribution by obtaining a distribution along the first direction; and (c2) when the second direction projection distribution is created, a second reference in each of the second direction projection distributions. The cumulative number of black pixels A process in which a continuous section along the first direction is defined as a second black pixel section, and a scanning region in which the second black pixel section is obtained is defined as a second divided row, and a divided line is cut out; A process of determining a line direction of the document image based on the first direction projection distribution and the second direction projection distribution when a line direction of the image is unknown; (e1) a line direction of the document image Is the first direction, wherein the divided rows included in the first direction divided areas adjacent to each other in the first direction are the first blacks obtained by scanning the divided rows. (E2) adjacent to each other in the second direction when the row direction of the document image is the second direction, and connecting the divided rows in which at least some of the pixel sections overlap each other. You The divided rows included in the second direction divided area are connected to each other, and at least a part of the second black pixel sections obtained by scanning the divided rows are overlapped with each other. A line segmentation process. 2. The line segmentation method according to claim 1, wherein in performing the process (d), the length of each of the first black pixel sections is defined as a first black pixel section length. A process of calculating the variation of the length of the black pixel section; a process of calculating the variation of the length of the second black pixel section by setting each length of the second black pixel section as a second black pixel section length; By comparing the variation in the length of one black pixel section with the variation in the length of the second black pixel section, the ratio between the variation in the length of the first black pixel section and the variation in the length of the second black pixel section is determined by a third reference value. A method for determining the scanning direction of the projection distribution from which the black pixel section length with the smaller variation is obtained as the row direction. 3. The line segmentation method according to claim 2, wherein in performing the process (d), the ratio of the variation in the first black pixel section length to the variation in the second black pixel section length is set to In a case where the distance is less than a third criterion, a length of a first white pixel section between the first black pixel sections adjacent to each other in the first direction projection distribution;
A line segmentation method including a process of determining a row direction using a length of a second white pixel section between adjacent second black pixel sections in a directional projection distribution. 4. The line segmentation method according to claim 1, wherein in performing the process (d), the first white pixel between the first black pixel sections adjacent to each other in the first direction projection distribution. The length of the section and the second
A line segmentation method including a process of determining a row direction using a length of a second white pixel section between adjacent second black pixel sections in a directional projection distribution. 5. The line segmentation method according to claim 3, wherein when the first white pixel section is one, the length of the first white pixel section is defined as a first average white pixel section length. And when there are a plurality of first white pixel sections, a process of calculating an arithmetic mean of the lengths of the first white pixel sections as a first average white pixel section length; When there is one pixel section, the length of the second white pixel section is set to the second average white pixel section length, and when there are a plurality of the second white pixel sections, the second white pixel section A process of calculating an arithmetic mean of each length of the section as a second average white pixel section length, and comparing the first average white pixel section length with the second average white pixel section length, Determining the scanning direction of the obtained projection distribution having a large average white pixel section length as the row direction. A line segmentation method characterized by the following. 6. The line segmentation method according to claim 1, wherein in performing the process (d), a process of obtaining a total number of the first direction divided regions as a first divided region number; A process of calculating the total number of black pixel sections over all of the first direction divided areas as a first black pixel section number; and a process of calculating the total number of the second direction divided areas as a second divided area number. A process of calculating the total number of the second black pixel sections over all of the second direction divided areas as a second black pixel section number; and the first black pixel section number and the first divided area number are: Matches,
A process of determining the row direction as the first direction when the number of the second black pixel sections does not match the number of the second divided areas; The number matches,
Determining a row direction as the second direction when the number of the first black pixel sections and the number of the first divided areas do not match. 7. The line segmentation method according to claim 1, wherein in the step (e1), a part of the first black pixel section of the first direction divided row included in the divided area is divided into the divided area. And a part of the first black pixel section of two first divided rows included in another adjacent divided area, the length of the overlapping section is longer than the fourth reference value. In the process (e2), a part of the second black pixel section of the second direction divided row included in the divided area is connected to another divided area adjacent to the divided area. In a case where each of the second black pixel sections of the two included second divided rows overlaps with a part of the second black pixel section, the divided rows whose overlapping section lengths are longer than a fifth reference value are connected. Line segmentation method. 8. The line cutting method according to claim 1, wherein the first (c1) process is performed before the (e1) process.
A process of calculating an arithmetic mean of the length of a first black pixel section (hereinafter also referred to as a first black pixel section length) of a divided row as a first average black pixel section length; The first black pixel section length is compared with the first black pixel section length, and the ratio of the first black pixel section length to the first average black pixel section length in the first black pixel section is smaller than a sixth reference value. A process of selecting a pixel section as a first candidate section; and a first candidate black pixel section that is the nearest first black pixel section on one side of the first candidate section and the first candidate section being one continuous A process for determining the length of the first integrated section, with the black pixel section as the first integrated section when the black pixel section is set, and the difference between the length of the first integrated section and the first average black pixel section length The size of the first candidate section, and the other of the first candidate section In the case where the second candidate black pixel section, which is the nearest first black pixel section on the side of, and the first candidate section are defined as one continuous black pixel section, the black pixel section is referred to as a second integrated section. A process of determining the length of the second integrated section; a process of determining the magnitude of the difference between the length of the second integrated section and the first average black pixel section length as a second integrated section length error; A process of comparing a long error with the second integrated section length error and selecting an integrated section having a smaller error as a first selected integrated section; and a first candidate section of the first selected integrated section. A process of calculating the length of a first white pixel section between the first white pixel section and the first black pixel section immediately adjacent to the first candidate section; comparing the length of the first white pixel section with a seventh reference value; When the length of the first white pixel section is shorter than the seventh reference value, the first selection is performed. Correcting the line cutout by setting the integrated section as one continuous first black pixel section. 9. The method according to claim 1, wherein before performing the process (e2), the second process is performed in the process (c2).
A process of calculating an arithmetic mean of the length of a second black pixel section (hereinafter, also referred to as a second black pixel section length) of a divided row as a second average black pixel section length; The second black pixel section length is compared with the second black pixel section length, and the ratio of the second black pixel section length to the second average black pixel section length in the second black pixel section is smaller than an eighth reference value. A process of selecting a pixel section as a second candidate section; and a third candidate black pixel section, which is the nearest second black pixel section on one side of the second candidate section, and the second candidate section A process for determining the length of the third integrated section, where the black pixel section is a third integrated section when the black pixel section is set, and a difference between the length of the third integrated section and the second average black pixel section length Calculating the size of the second candidate section as a third integrated section length error; When the fourth candidate black pixel section, which is the nearest second black pixel section on the other side, and the second candidate section are defined as one continuous black pixel section, the black pixel section is referred to as a fourth integrated section. A process of calculating the length of the fourth integrated section; a process of calculating the difference between the length of the fourth integrated section and the second average black pixel section length as a fourth integrated section length error; Comparing the integrated section length error with the fourth integrated section length error and selecting the integrated section having the smaller error as the second selected integrated section; and selecting the second selected integrated section of the second selected integrated section. A process of determining the length of the second white pixel section between the candidate section and the second black pixel section immediately adjacent to the second candidate section; and comparing the length of the second white pixel section with the ninth reference value. When the length of the second white pixel section is shorter than the tenth reference value, Correcting the line cutout by setting the second selected integrated section as one continuous second black pixel section. 10. The line cutout method according to claim 1, wherein the first cutout in the process (c1) is performed.
A process of calculating an arithmetic mean of the length of a first black pixel section (hereinafter, also referred to as a first black pixel section length) of a divided row as a first average black pixel section length; The first black pixel section length is compared with the first black pixel section length, and the ratio of the first black pixel section length to the first average black pixel section length in the first black pixel section is greater than an eleventh reference value. A process of selecting a pixel section as a first division candidate section; and a step of selecting a pixel section as a first division candidate section along the first direction at a position where a black pixel is detected on a scanning line that scans a center of the first division candidate section. A process for obtaining a position closest to the other side as a first position, a process for obtaining a first position closest to the other side along the first direction as the closest first position among the first positions, and the scanning line. Run around the center of the first division candidate section A process of determining, as a second position, a position closest to the other side along the second direction among positions where black pixels are detected on a scanning line to be inspected; and a process along the second direction among the second positions. A process for determining the second position closest to the one side as the closest second position, comparing the closest first position with the closest second position, and determining that the closest first position is the closest second position. When the second position is closer to the other side than the second position and when the nearest first position and the nearest second position are located on the same straight line along the second direction, the first
A process of correcting the line segmentation by dividing the division candidate section into two in the center of the first division candidate section to generate two divided black pixel sections; Comparing the nearest second position with the nearest second position, if the nearest first position is closer to the one side than the nearest second position, the intermediate position between the nearest first position and the nearest second position; And dividing the first candidate division section with a division line along the second direction, thereby generating the first one division area on one side and the other division area on the other side, and The nearest second position is divided by the obtained scanning line to generate two divided black pixel sections, and the other divided area is divided by the scanning line from which the nearest first position is obtained. Process of correcting line segmentation by generating two divided black pixel sections And a line segmentation method. 11. The line cutting method according to claim 1, wherein the second line cut in the process (c2).
A process of calculating an arithmetic mean of the length of a second black pixel section (hereinafter, also referred to as a second black pixel section length) of a divided row as a second average black pixel section length; The second black pixel section length is compared with the second black pixel section length, and the ratio of the second black pixel section length to the second average black pixel section length in the second black pixel section is larger than a twelfth reference value. Selecting a pixel section as a second division candidate section; and selecting a pixel section as a second division candidate section along the second direction at a position where a black pixel is detected on a scanning line that scans the vicinity of the center of the second division candidate section. A process of obtaining a position closest to the other side as a third position, a process of obtaining a third position closest to the other side along the second direction among the third positions as a closest third position, Of the second division candidate section A process of determining a position closest to the other side along the first direction as a fourth position among positions where black pixels are detected on a scanning line to be inspected, and a process along the first direction among the fourth positions. A process of determining the fourth position closest to the one side as the closest fourth position, and comparing the third closest position with the fourth closest position to determine that the third position is the closest fourth position. When the second division candidate section is located on the other side than the four positions, the second division candidate section is divided into two at the center of the second division candidate section to generate two divided black pixel sections, whereby line segmentation is performed. Comparing the closest third position and the closest fourth position with each other to determine whether the third closest position is closer to the one side than the fourth closest position and When the third position and the nearest fourth position are located on the same straight line along the first direction. Dividing the second division candidate section by a division line passing through an intermediate point between the third nearest position and the fourth nearest position and along the first direction, whereby the first one of the first sides is divided. Generating a divided region and the other divided region on the other side, dividing the one divided region by a scanning line from which the nearest fourth position is obtained, to generate two divided black pixel sections, and Correcting the line segmentation by dividing the region by the scanning line from which the closest third position is obtained to generate two divided black pixel sections. 12. The document image is scanned in a first direction substantially along a line direction of the document image, and a cumulative number of black pixels is detected for each scan. A projection distribution is created by calculating a distribution along a second direction orthogonal to the following. In the projection distribution, each black pixel section in which a projection portion having a cumulative number of black pixels equal to or greater than a thirteenth reference value is continuous in the second direction is After performing line segmentation with each of the obtained scanning regions as one line region, when performing line segmentation correction, the length of the black pixel section of the segmented line area (hereinafter also referred to as black pixel section length). Calculating the arithmetic mean of the average black pixel section length as the average black pixel section length; and comparing the average value with the black pixel section length to determine that the ratio of the black pixel section length to the average value in the black pixel section is 14th. Black pixel area smaller than the reference value And selecting the interval as a candidate section, and selecting the first candidate black pixel section, which is the nearest black pixel section on one side of the candidate section, and the black when the candidate section is regarded as one continuous black pixel section. Setting a pixel section as a first integrated section and calculating a length of the first integrated section; calculating a difference between the length of the first integrated section and the average value as a first integrated section length error; When the second candidate black pixel section, which is the nearest black pixel section on the other side of the candidate section, and the candidate section are defined as one continuous black pixel section, the black pixel section is referred to as a second integrated section. A process of calculating the length of the second integrated section; a process of calculating a difference between the length of the second integrated section and the average black pixel section length as a second integrated section length error; By comparing the error with the second integrated section length error, A process of selecting a smaller integrated section as a selected integrated section; and a process of determining a length of a white pixel section between the candidate section and the nearest black pixel section of the candidate section in the selected integrated section; The length of the white pixel section is compared with a fifteenth reference value. If the length of the white pixel section is shorter than the fifteenth reference value, the selected integrated section is regarded as one continuous black pixel section. And a process for correcting the line segmentation. 13. A scanning line for scanning the document image by setting a first direction substantially along a line direction of the document image as a main scanning direction and a second direction orthogonal to the first direction as a sub-scanning direction. A projection distribution is created by detecting the cumulative number of black pixels for each scan, and a black pixel section in which the projected portion of the cumulative black number of pixels equal to or larger than a 16th reference value is continuous in the second direction in the projection distribution is obtained. After performing line segmentation with each region as one line region, when performing line segmentation correction, the average value of the lengths of black pixel sections (hereinafter, also referred to as black pixel section lengths) of the segmented line areas. And comparing the average value with the black pixel section length to divide the black pixel section of the black pixel section in which the ratio of the black pixel section length to the average value is greater than a seventeenth reference value. Select as a candidate section And a process in which a black pixel is detected on a scan line that scans the vicinity of the center of the division candidate section in the scan line.
A process of determining a position closest to one side along a direction as a first position; a process of obtaining a first position closest to the other side along the first direction among the first positions as a first position; Among positions where black pixels are detected on a scanning line that scans the vicinity of the center of the division candidate section in the scanning line, the second
A process of obtaining the position closest to the other side along the direction as a second position, and a process of obtaining the second position closest to the side along the second direction among the second positions as the closest second position Comparing the nearest first position with the nearest second position, and determining whether the nearest first position is on the other side of the nearest second position, and When the nearest second position is located on the same straight line along the second direction, the division candidate section is divided into two at the center of the division candidate section to form two divided black pixel sections. Generating a line segmentation correction process, comparing the nearest first position with the nearest second position, and comparing the nearest first position with the nearest second position, Side, passes through the middle between the nearest first position and the nearest second position, and By dividing the first division candidate section by a division line along the direction, the one divided region on the one side and the other divided region on the other side are generated, and the one second divided region is defined by the nearest second position. Two divided black pixel sections are generated by dividing by the obtained scanning line, and two divided black pixel sections are generated by dividing the other divided area by the scanning line from which the nearest first position is obtained. And performing a process of correcting the line cutout. 14. When the line direction of the document image is unknown and when the line direction of the document image is the first direction, the document image is divided by a dividing line along a second direction orthogonal to the first direction. A first direction divided region generating unit that generates a plurality of first direction divided regions arranged along the first direction by dividing the first direction divided region. Scanning each of the direction divided areas along the first direction,
A first direction projection distribution creation unit for creating a first direction projection distribution by detecting a cumulative number of black pixels for each scan and obtaining a distribution of the cumulative number of black pixels along the second direction; When the projection distribution is created, in each of the first direction projection distributions, a section in which the cumulative number of black pixels equal to or greater than the first reference value is continuous along the second direction is defined as a first section.
As a black pixel section, a first divided line cutout unit that performs a divided line cutout using a scanning area in which the first black pixel section is obtained as a first divided line, and a case where the line direction of the document image is unknown and the document image When the row direction is a second direction, by dividing the document image by a dividing line along the first direction, a plurality of second direction divisions arranged along a second direction orthogonal to the first direction A second direction divided area generating unit that generates an area; and, when the second direction divided area is generated, scanning each of the second direction divided areas along the second direction.
A second direction projection distribution creating unit that detects a cumulative number of black pixels for each scan and obtains a distribution of the cumulative number of black pixels along the first direction to create a second direction projection distribution; When the projection distribution is created, in each of the second direction projection distributions, a section in which the cumulative number of black pixels equal to or more than the second reference value is continuous along the first direction is defined as a second section.
As a black pixel section, a second divided line cutout unit that performs a divided line cutout using a scanning area in which the second black pixel section is obtained as a second divided line, and, when the line direction of the document image is unknown, A line direction determining unit configured to determine a line direction of the document image based on a first direction projection distribution and the second direction projection distribution, wherein the line direction of the document image is the first direction; The divided rows included in the first direction divided areas adjacent to each other in one direction, wherein the divided rows at least partially overlap the first black pixel section obtained by scanning the divided rows. A first divided line connecting unit that connects lines to each other to separate lines, and includes, when the line direction of the document image is the second direction, the second direction divided regions adjacent to each other in the second direction. Said A second divided row connection unit that connects divided rows in which at least a part of the second black pixel section obtained by scanning the divided row overlaps with each other, and cuts out the rows. A line cutting device characterized by being provided. 15. The line segmenting device according to claim 14, wherein the row direction determination unit sets each length of the first black pixel section as a first black pixel section length, the first black pixel section. A first variation calculation unit that calculates a variation in length; and a second variation calculation that calculates a variation in the length of the second black pixel section by setting each length of the second black pixel section as a second black pixel section length. And comparing the variation of the first black pixel section length with the variation of the second black pixel section length to determine a ratio between the variation of the first black pixel section length and the variation of the second black pixel section length. And a variation comparison / determination unit that determines the scanning direction of the projection distribution from which the black pixel section length with smaller variation is obtained as the row direction when is equal to or greater than the third reference value. apparatus. 16. The line segmentation device according to claim 15, wherein the row direction determination unit determines that a ratio of the variation of the first black pixel section length to the variation of the second black pixel section length is the third reference. If the value is less than the value, the length of the first white pixel section between the first black pixel sections adjacent to each other in the first direction projection distribution;
A row segmentation apparatus, wherein a row direction is determined using a length of a second white pixel section between adjacent second black pixel sections in a directional projection distribution. 17. The line segmenting device according to claim 14, wherein the line direction determination unit is configured to determine a first white pixel section between the first black pixel sections adjacent to each other in the first direction projection distribution. The length of the second
A row segmentation apparatus, wherein a row direction is determined using a length of a second white pixel section between adjacent second black pixel sections in a directional projection distribution. 18. The line segmentation device according to claim 16, wherein the row direction determination unit determines the length of the first white pixel section when the number of the first white pixel section is one. If one average white pixel section length is used and the first white pixel section is plural, an arithmetic average of the respective lengths of the first white pixel section is calculated as the first average white pixel section length. A first average white pixel section length calculation unit, wherein, when the number of the second white pixel sections is one, the length of the second white pixel section is set to a second average white pixel section length; When there are a plurality of pixel sections, a second average white pixel section length calculation unit that calculates an arithmetic mean of the lengths of the second white pixel sections as a second average white pixel section length; The average white pixel section length is compared with the second average white pixel section length to obtain a white pixel section having a larger average value. An average value comparing and judging unit for judging the scanning direction of the projection distribution obtained as the row direction. 19. The line segmentation device according to claim 14, wherein the line direction determining unit calculates a total number of the first direction divided regions as a first divided region number, and a first divided region number counting unit. A first black pixel section number counting unit that calculates a total number of the first black pixel sections over all the first direction divided areas as a first black pixel section number; and a total number of the second direction divided areas. As the number of second divided areas, and a second black section that calculates the sum of the number of the second black pixel sections over all of the second direction divided areas as the number of second black pixel sections. A pixel section number counting unit, wherein the first black pixel section number and the first divided area number match,
And when the number of the second black pixel sections does not match the number of the second divided regions, the row direction is determined to be the first direction,
Alternatively, when the number of the second black pixel sections and the number of the second divided areas match and the number of the first black pixel sections does not match the number of the first divided areas, the row direction is changed to the second direction. A line segmentation device comprising: a section number division number comparison determination unit that determines two directions. 20. The line segmenting device according to claim 14, wherein the first divided row connecting unit is configured to determine that a part of the first black pixel section of the first direction divided row included in the divided area is divided by the divided area. When each of the first black pixel sections of two first divided rows included in another divided area adjacent to the area overlaps with a part of the first black pixel section, the length of the overlapping section is larger than the fourth reference value. Connecting the long divided rows with each other, wherein the second divided row connecting unit is configured such that a part of the second black pixel section of the second direction divided row included in the divided area is another one adjacent to the divided area. When overlapping with a part of the second black pixel section of the two second divided rows included in the divided area, connecting the divided rows whose overlapping section is longer than the fifth reference value. A line cutting device characterized by the above-mentioned. 21. The line segmenting apparatus according to claim 14, further comprising a first integrated correcting unit for correcting line segmenting, wherein the first integrated modifying unit is segmented by the first divided line segmenting unit. First
A first average black pixel section length calculation unit that determines an arithmetic mean of the length of the first black pixel section (hereinafter, also referred to as a first black pixel section length) of the divided row as the first average black pixel section length; Comparing the first average black pixel section length with the first black pixel section length, the ratio of the first black pixel section length to the first average black pixel section length in the first black pixel section is the sixth. A first candidate section selection unit that selects a first black pixel section smaller than the reference value as a first candidate section; and a first candidate black pixel that is a first black pixel section closest to one side of the first candidate section. When the section and the first candidate section are set as one continuous black pixel section, the black pixel section is defined as a first integrated section, and a first integrated section length detecting unit that determines the length of the first integrated section; The magnitude of the difference between the length of the first integrated section and the first average black pixel section length is defined as a first integrated section length error. A first integrated section length error calculating unit to calculate, a second candidate black pixel section that is the nearest first black pixel section on the other side of the first candidate section, and the first candidate section as one continuous black A second integrated section length detecting unit for determining the length of the second integrated section, wherein the black pixel section in the case of the pixel section is a second integrated section; and the length of the second integrated section and the first average black pixel A second integrated section length error calculating unit that calculates the magnitude of the difference from the section length as a second integrated section length error, and comparing the first integrated section length error and the second integrated section length error, A first integrated section selection unit that selects an integrated section having a smaller one as a first selected integrated section; and a first candidate section of the first selected integrated section and a first first integrated section closest to the first candidate section. A first white pixel section length detection unit for determining the length of the first white pixel section between the black pixel section and The length of the first white pixel section is compared with a seventh reference value, and when the length of the first white pixel section is shorter than the seventh reference value, the first selected integrated section is connected to one continuous The first black pixel section is used to correct the line segmentation.
A line segmentation device comprising an integration unit. 22. The line segmenting apparatus according to claim 14, further comprising a second integrated correcting unit that corrects line segmenting, wherein the second integrated modifying unit is segmented by the second divided line segmenting unit. Second
A second black pixel section length average value calculation unit that obtains an arithmetic mean of the length of the second black pixel section of the divided row (hereinafter, also referred to as a second black pixel section length) as a second average black pixel section length; Comparing the second average black pixel section length with the second black pixel section length, the ratio of the second black pixel section length to the second average black pixel section length in the second black pixel section is equal to the second average black pixel section length. A second candidate section selection unit that selects a second black pixel section smaller than 8 reference values as a second candidate section, and a third candidate black that is a second black pixel section closest to one side of the second candidate section. When the pixel section and the second candidate section are set as one continuous black pixel section, the black pixel section is defined as a third integrated section, and a third integrated section length extracting unit for calculating the length of the third integrated section is provided. The magnitude of the difference between the length of the third integrated section and the length of the second average black pixel section is determined by a third integrated section length error. A third integrated section length error calculating unit that calculates the fourth candidate black pixel section that is the nearest second black pixel section on the other side of the second candidate section and the second candidate section The black pixel section when the black pixel section is obtained as a fourth integrated section, a fourth integrated section length extracting unit for calculating the length of the fourth integrated section, the length of the fourth integrated section and the second average A fourth integrated section length error calculating unit that calculates the magnitude of the difference from the black pixel section length as a fourth integrated section length error, and comparing the third integrated section length error with the fourth integrated section length error A second selected integrated section selecting unit that selects the integrated section having a smaller error as the second selected integrated section; and, among the second selected integrated sections, the second candidate section and the nearest second candidate section. A second white pixel section length extraction unit for determining the length of the second white pixel section between the second black pixel section and The length of the second white pixel section is compared with a ninth reference value, and if the length of the second white pixel section is shorter than the tenth reference value, the second selected integrated section is defined as one continuous A second integration section that corrects line segmentation by setting the second black pixel section as the segment. 23. The line segmentation device according to claim 14, further comprising a first division modification unit for modifying line segmentation, wherein the first division modification unit is configured to cut out the first segmented line segmentation unit. First
A first average black pixel section length calculation unit that determines an arithmetic mean of the length of the first black pixel section of the divided row (hereinafter, also referred to as a first black pixel section length) as the first average black pixel section length; Comparing the first average black pixel section length with the first black pixel section length, the ratio of the first black pixel section length to the first average black pixel section length in the first black pixel section is eleventh. A first division candidate section selection unit that selects a first black pixel section larger than a reference value as a first division candidate section; and a black pixel on a scan line that scans the center of the first division candidate section among the scan lines. Among the detected positions, a first position extracting unit that determines a position closest to the one side along the first direction as a first position, and a position closest to the other side along the first direction among the first positions. A nearest first position extracting unit that determines the first position of the first position as the nearest first position; Among the positions where black pixels are detected on a scanning line that scans the vicinity of the center of the first division candidate section among the scanning lines, a position closest to the other side along the second direction is determined as a second position. A two-position extracting unit; and, of the second positions, a nearest second position that determines the closest second position along the second direction as a nearest second position.
A position extracting unit, comparing the nearest first position and the nearest second position, and when the nearest first position is on the other side of the nearest second position, and When the first position and the nearest second position are located on the same straight line along the second direction, the first position
A first division unit that corrects line segmentation by dividing the division candidate section into two in the center of the first division candidate section to generate two divided black pixel sections; And comparing the nearest second position with the nearest second position. If the nearest first position is closer to the one side than the nearest second position, the nearest first position and the nearest second position are compared with each other. By dividing the first division candidate section by a division line passing through the middle of the first direction and along the second direction, thereby generating the first first division area on one side and the other division area on the other side, The divided region is divided by the scanning line from which the nearest second position is obtained to generate two divided black pixel sections, and the other divided region is divided by the scanning line from which the nearest first position is obtained. To correct the line segmentation by generating two divided black pixel sections A line cutting device comprising a split part. 24. The line segmentation apparatus according to claim 14, further comprising a second division modification unit that modifies the line division, wherein the second division modification unit is configured to extract the line segmented by the second divided line segmentation unit. Second
A second average black pixel section length calculation unit that calculates an arithmetic average of the length of the second black pixel section of the divided row (hereinafter, also referred to as a second black pixel section length) as the second average black pixel section length; Comparing the second average black pixel section length with the second black pixel section length, the ratio of the second black pixel section length to the second average black pixel section length in the second black pixel section is 12th. A second division candidate section selection unit that selects a second black pixel section larger than a reference value as a second division candidate section; and a black pixel on a scan line that scans near the center of the second division candidate section among the scan lines. Among the detected positions, a third position extracting unit that determines a position closest to the one side along the second direction as the third position, and a position closest to the other side along the second direction among the third positions. A third closest position extracting unit that determines the third position of the second position as the third closest position; Of the positions where black pixels are detected on a scanning line that scans the vicinity of the center of the second division candidate section among the scanning lines, the position closest to the other side along the first direction is determined as a fourth position. A four-position extracting unit, and, among the fourth positions, a fourth fourth position that is closest to the one side along the first direction and is determined as a fourth closest position.
A position extracting unit that compares the third closest position with the fourth closest position and, if the third closest position is closer to the other side than the fourth closest position, A third division unit that corrects line segmentation by dividing the division candidate section into two at the center of the second division candidate section to generate two divided black pixel sections; And the nearest fourth position, and when the nearest third position is on the one side of the nearest fourth position, and when the nearest third position and the nearest fourth position are different from each other. When located on the same straight line along the first direction, the second division candidate passes through an intermediate point between the nearest third position and the nearest fourth position and is divided along the first direction. By dividing a section, the first divided region on the one side and the other divided region on the other side are generated, and the one divided region is The second divided position is divided by the scanning line at which the fourth closest position is obtained to generate two divided black pixel sections, and the other divided region is divided by the scanning line at which the third closest position is obtained to generate two divided black pixel sections. A line segmentation device comprising: a fourth segmentation unit that corrects line segmentation by generating a divided black pixel section. 25. The line segmenting apparatus according to claim 14, wherein it is determined whether or not information indicating a line direction of the document image is added to information of a document image externally input to the line segmenting device. If the information indicating that the line direction is the first direction is added, the information of the document image is transmitted to the first divided area generation unit, and the information that the line direction is the second direction is indicated. When the information is added, the information of the document image is transmitted to the second divided area generation unit, or
When the information indicating the line direction is non-attached, the line direction confirmation switching unit that sends the information of the document image to the first and second divided region generation units; If the direction is determined to be one direction, the information obtained from the first divided line segmentation unit is output. If the line direction is determined to be the second direction, the second segmented line segmentation is performed. And a switching output unit for outputting information obtained from the unit. 26. The document image is scanned in a first direction substantially along a row direction of the document image to detect an accumulated number of black pixels for each scan, and the accumulated number of black pixels in the first direction is detected. And a distribution along a second direction orthogonal to the above is created to form a projection distribution, and in the projection distribution, a projected portion having a cumulative number of black pixels equal to or larger than a thirteenth reference value is continuous in each black pixel section in the second direction. Is a line cutout correction device that corrects the line cutout obtained by performing line cutout with each of the scan regions obtained as one line region, wherein the length of the black pixel section of the cutout line region (hereinafter, referred to as An average black pixel section length calculation unit for calculating an arithmetic mean of the black pixel section length as an average black pixel section length; and comparing the average black pixel section length with the black pixel section length to obtain the black pixel section length. Of the average black pixel section length A candidate section selection unit that selects a black pixel section whose pixel section length ratio is smaller than the fourteenth reference value as a candidate section; and a first candidate black pixel section that is the nearest black pixel section on one side of the candidate section. A first integrated section length extraction unit that determines the black pixel section as a first integrated section when the candidate section is set as one continuous black pixel section; A first integrated section length error calculation unit that calculates the magnitude of the difference between the section length and the average black pixel section length as a first integrated section length error; When a certain second candidate black pixel section and the candidate section are regarded as one continuous black pixel section, the black pixel section is defined as a second integrated section, and a second integrated section length for calculating the length of the second integrated section An extraction unit, which calculates a length of the second integrated section and the average black pixel section length A second integrated section length error calculating section that calculates the size of the second integrated section length error as a second integrated section length error, comparing the first integrated section length error and the second integrated section length error, A selection integration section selection unit that selects a section as a selection integration section; and a white pixel section that calculates the length of a white pixel section between the candidate section of the selection integration section and the nearest black pixel section of the candidate section. A length extracting unit, comparing the length of the white pixel section with a fifteenth reference value, and, when the length of the white pixel section is shorter than the fifteenth reference value, combining the selected integrated section with one continuous A line cutout correction device, comprising: an integration unit that corrects line cutouts by setting black pixel sections. 27. Scanning the document image along a first direction substantially along a line direction of the document image, detecting a cumulative number of black pixels for each scan, and detecting the cumulative number of black pixels in the first direction. A projection distribution is created by calculating a distribution along a second direction orthogonal to the above, and in the projection distribution, a black pixel section in which a projection portion having a cumulative number of black pixels equal to or greater than a 16th reference value is continuous in the second direction is defined. What is claimed is: 1. A line cutout correction device for correcting a line cutout obtained by performing line cutout with each obtained scanning region as one line region, wherein the length of a black pixel section of the cutout line region (hereinafter referred to as black) An average black pixel section length calculation unit for calculating an average black pixel section length of the black pixel section length, and comparing the average black pixel section length with the black pixel section length. Ratio of black pixel section length to average black pixel section length A division candidate section selection unit that selects a black pixel section larger than a seventeenth reference value as a division candidate section; and a position where a black pixel is detected on a scanning line that scans the center of the division candidate section among the scanning lines. Among the first
A first position extracting unit that determines a position closest to one side along the direction as a first position, and among the first positions, a first position closest to the other side along the first direction as a closest first position. A nearest first position extracting unit to be obtained; and a second position among black pixels detected on a scanning line that scans the center of the division candidate section in the scanning line.
A second position extraction unit that determines the position closest to the other side along the direction as a second position; and a second position closest to the one side along the second direction among the second positions. Second nearest location
A position extracting unit, comparing the nearest first position and the nearest second position, and when the nearest first position is on the other side of the nearest second position, and When one position and the nearest second position are located on the same straight line along the second direction, the division candidate section is divided into two at the center of the division candidate section, and two divided black sections are divided. By generating a pixel section, a first division unit that corrects line segmentation is compared with the nearest first position and the nearest second position, and the nearest first position is determined by the nearest first position. Dividing the first candidate segment by a dividing line that passes through the middle between the nearest first position and the nearest second position and extends in the second direction when the first position is closer to the one side than the second position; Thus, the one divided region on the one side and the other divided region on the other side are generated, and the one divided region is And the other divided area is divided by the scanning line from which the nearest first position has been obtained to generate two divided black pixel sections. And a second dividing unit that corrects the line cutout by generating a line cutout.