JPH08305794A - Address line extracting device - Google Patents

Abstract

PURPOSE: To provide an address line extracting device capable of highly accurately extracting an address line without being influenced by change in the size of address characters. CONSTITUTION: A labeling part 22 labels the image of an area including an address to obtain blocks. A block sorting part 23 locally groups these blocks and a character size estimating part 24 calculates reference block size to be the reference size of blocks in each group. A base block selecting part 26 selects a base block to be a block starting the integration of blocks. At the time of judging that an integrated block generated by integrating blocks from the base block by a line growing part 27 is probably an address line, a line inspecting part 29 outputs the integrated block, and at the time of judging that the integrated block is not probably an address line, allows respective parts to execute the reselection of a base block and the regeneration of an integrated block. When a base block can not be detected, the calculation of reference block size is retried.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an address line extraction device, and more particularly to an address line extraction device for extracting address lines of a mail.

[0002]

2. Description of the Related Art In a document image recognition apparatus, a layout analysis process is performed on a digital image obtained by scanning a document, and then a character recognition process is performed on each extracted character image to obtain individual characters. The process proceeds in a procedure of obtaining a character code according to the image. Here, the layout analysis process refers to a process from extracting a character line from an input image to performing format determination. As a document related to the layout analysis process, for example, in 1985, the Institute of Electronics, Information and Communication Engineers (Vol.69-D No.11), pages 2123 to 2241 published by Babaguchi et al. Entitled "Basic Study of Character Extraction from Handwritten Character Strings".

[0003] This paper discloses a technique for layout analysis of a mailing address, and the layout analysis is performed according to the procedure described below.

Isolated point removal is performed on a digital image obtained by scanning the surface of a mail piece. 8-connected labeling is performed to obtain a circumscribed rectangle of connected components. If the areas of the overlapping portions of the obtained rectangles are equal to or larger than the threshold value, these rectangles are integrated. The center coordinates of the rectangle are weighted by the rectangle area and projected in the vertical and horizontal directions, and the point having a high degree of concentration of the projected points is set as the address line position. If the rectangles after rectangle integration are close to squares, the combination of all rectangles at the minimum distance is performed. The average value of the rectangular areas is defined as the standard rectangular area. When the aspect ratio of the rectangle is equal to or larger than the default value, the image in the rectangle is projected onto the long side of the rectangle, and the rectangle is divided by the minimum value. The rectangles are integrated in the character string direction to generate address line candidates.

Thus, in this technique, the connected components obtained by the labeling process are integrated to obtain
We are trying to extract address lines.

[0006]

In the technology for detecting an address character by integrating connected components, which is represented by the above-mentioned technology, there are many cases where the size of the address character is fixed regardless of the postal matter. Was susceptible to fluctuations in size. Also, when separating line-to-line contact, a fixed value is used as the character size in order to calculate the separation point, which is a cause of degrading the accuracy of the cutout position.

Generally, when the address is handwritten,
There are often significant variations in the font size that is printed on a piece of mail. Therefore, unlike many methods up to now, it is not expected that the character size will change from character to character, and the technology that extracts addressing lines assuming a single character size does not It was not possible to cope with variations in character size, and there were cases where the wrong address line was extracted.

In the technique proposed in the "basic consideration of character segmentation from a handwritten character string" described above, a standard size of a character is obtained, and it is attempted to separate contact between characters based on this standard size. There is. However, since the case where miscellaneous information other than address writing is mixed on the surface of the mail is not considered, it is impossible to estimate the character size with high accuracy. As a result, the extracted address lines were also unstable and uncertain.

In actual mail, lines may be intricate or contact with each other, but no consideration has been given to that. Therefore, the above-described technique has a problem in that in such a case, it is not possible to estimate the standard character size and extract complicated lines such as separation of line-to-line contact.

Therefore, an object of the present invention is to provide an address line extracting device which can accurately extract an address line regardless of the character size used in the address.

Another object of the present invention is to provide an address line extracting device which can accurately extract an address line even with complicated address lines.

[0012]

According to the first aspect of the present invention,
(B) miscellaneous information detection means for detecting information other than the address from the image data according to the description content of the mail, and (b) use of the detection result of this miscellaneous information detection means to describe the address of the mail. And (c) for each format candidate assumed by this address format assumption means, which assumes at least one format candidate including information specifying the writing direction of the address line A block specifying unit that specifies a rectangular block having a minimum size including a closed area in which black pixels are connected from image data for one area corresponding to a format candidate and outputs information about the specified block,
(D) Block classification means for classifying each block into several groups for each format candidate based on the information on the position and size of each block specified by this block identification means for each format candidate, ) For each group classified by this block classification means, a standard block size calculation means for calculating an average value of the sizes of blocks belonging to the groups and outputting the calculation result as a standard block size,
(F) Blocks that belong to each group classified by the block classifying unit and whose difference in size from the standard block size calculated for the group by the standard block size calculating unit is within a predetermined value are addressed character blocks. The address character block identification means for outputting the identification result for each group and (g) the address character block identified for each group by this address character block identification means has a size within a predetermined range. , A base block selecting means for selecting as the base block the address character block having the size closest to the standard block size for that group, and (h) the base block selected for each group by this base block selecting means, and its group. Orthogonal to the writing direction in the format candidate corresponding to The degree of overlap with respect to direction to integrate the address character blocks in the same group of more than a predetermined value,
A block integrating means for outputting the integrated result for each group as an integrated block, and (b) this block integrating means mutually determines the position and size in the image data of the integrated block output for each group in units of format candidates. A format determining unit that compares the format candidates for which the integrated block having the most suitable form as the address line is obtained to the format used in the mail.

That is, according to the first aspect of the present invention, several forms are assumed for the postal matter to be processed, and the forms are used for each assumed form. First, the block corresponding to one character is obtained for the image data of one area related to the portion where the address is likely to exist, and the blocks are roughly divided into groups. Then, after grouping, for each group, the average size of the characters included in the group is estimated, and the estimated line size (standard block size) is used to integrate the address lines ( The address line extraction device is configured so that the integrated block) is generated.

As a result, when the character blocks are integrated and the address line is extracted, the base block to be the block for starting the integration is selected according to the size of each character block. As a result, the precision of the address line obtained as a result of the block integration is achieved. When calculating the standard block size,
The length of the long side of the block may be used, or the square root of the area of the block or the width of the block may be used.

Further, as in the invention described in claim 2, the standard block size calculation means in the invention described in claim 1 has a size outside the range of the predetermined upper limit value and lower limit value from the character blocks belonging to the group. Exclude the block, calculate the average value and standard deviation of the size of the blocks in the remaining group, and the difference between the calculated average value and the size If the average value is calculated again and the calculation result is output as the standard block size, the base block will be selected more accurately, and as a result,
It is possible to extract address lines with high accuracy.

According to the third aspect of the invention, the addressed character block specifying means calculates the block belonging to each group classified by the block classifying means for the group by the standard block size calculating means. Blocks whose size difference from the standard block size is within a specified value are identified as address character blocks, and
If there is a block whose size is larger than the standard block size by a specified value or more, divide the block into blocks with the same size as the standard block size, and specify the divided blocks as address character blocks. If you configure the address line extraction device to
It is also possible to extract an accurate address line from a postal matter in which the address line is complicated.

As the block integrating means, the integrated block obtained as an integrated result is an upper limit value and a lower limit value of the width and height determined by a predetermined number of times the standard block size. It is determined whether or not the condition of being included in the range is satisfied, and when it is determined that the condition is not satisfied, the integrated block is discarded, and the base block selecting means selects from the character blocks belonging to the group, If the address line extraction device is configured using a block that is not yet used as a base block, and that selects the smallest block whose size difference from the standard block size is within a predetermined value as the base block. Can further reduce the possibility that the wrong address line is extracted.

Further, as in the invention described in claim 5, when there is no block that can be a base block in a certain group, the base block selecting means notifies the standard block size calculating means to the address line in the group. When the standard block size is calculated using unintegrated blocks, two lines with different character sizes cannot be separated into different groups in advance by the block classification means. In addition, since the correct average value of the block size is calculated, the address lines can be accurately extracted even when there is contact between address lines having different character sizes. .

[0019]

EXAMPLES The present invention will be described in detail below with reference to examples.

First embodiment

FIG. 1 is a functional block diagram of an address line extraction device according to the first embodiment of the present invention. As shown in the figure, the address line extraction device of the first embodiment is provided with an image input unit 11
A miscellaneous information detection unit 12, an address format presumption unit 13, a plurality of address line extraction units 14 _{1 to} 14 ₃ and a format determination unit 15 are provided.
Each unit is formed using a known electronic element such as a processor and a memory, and operates as described below.

The image input unit 11 scans a mail item and outputs a multivalued digital image corresponding to an image of a character or a postal code frame drawn on the surface thereof. Miscellaneous information detector 1
2 removes noise from the digital image output by the image input unit 11. Then, based on the digital image from which noise has been removed, the position and size of objects other than the address, such as stamps and postal code frames, etc., on the surface of the mail are detected, and the detection results are used as address peripheral information. Is output to the address format assumption unit 13.

The address format assumption unit 13 is an address format assumption unit 1.
Based on the address peripheral information from 3, the format of postal matter to be processed is determined (supposed) and the result is output.

The operation contents of the address format presumption unit will be described with reference to FIG. As shown in the figure, when writing an address on a postal matter, the longitudinal direction is different from the case where the postal matter is placed (longitudinal) so that the longitudinal direction is the vertical direction (a, b). There is a case (c) when the postal matter is placed so that it is in the left-right direction (placed horizontally), and in the writing direction of the letters,
When writing characters vertically (a) and horizontally (b,
There is c).

The address format presumption unit 13 determines the direction in which the mail is scanned based on the address peripheral information (for example, the position of the stamp, the position of the postal code frame) from the miscellaneous information detection unit,
When a line segment is detected in the image, the writing direction of the character is limited from the direction of the line segment, and the mail item whose address line is to be extracted is shown in FIG. It is determined (assumed) which one of the formats shown in (1) is used, and the result is output to the address line extraction unit 14. In the following description, the format shown in FIG. 2A is "vertical placement and vertical writing", the format such as (b) is "vertical placement and horizontal writing", and the format such as (c) is "horizontal placement." Horizontal writing "
Will be written as.

That is, when the line segment parallel to the longitudinal direction is extracted, the address format presuming unit 13 limits the format to "vertical placement vertical writing" or "horizontal placement horizontal writing". When it is not possible (when the formats cannot be narrowed down uniquely), multiple format candidates that may be used as those formats are output. Further, when the format candidates are output, information on an area where an address is likely to exist (hereinafter referred to as an address existence area) is also output.
For example, in the case of "vertical writing", the address is likely to exist in the right half from the bottom of the postal code frame to the end of the mail when the stamp is placed on the upper side. The area is set, and information about the set address existence area is output together with the format candidate.

Each address line extraction unit 14 receives the format candidate output by the address format assumption unit 13 and the coordinate information of the address existence area corresponding to the format, and generates the address line according to the format. Output as address line. Details of the operation of the address line extraction unit will be described later. Address format assumption section 1
When a plurality of format candidates are output from 3, the plurality of address line extraction units 14 are activated, and each address line extraction unit 14 executes address line extraction processing according to different format candidates.

The format determining unit 15 receives the address line candidates extracted by each address line extracting unit 14 as input, determines the certainty of each address line candidate, identifies the most probable format, and outputs the identified format.

The operation of the format determining section will be described in detail below with reference to FIG. Note that FIG. 3 is a diagram showing an example in which two address line extraction units are activated in the address line extraction device of the first embodiment. In FIG. 3A, address line extraction is performed. The images shown are shown in (b) and (c).
The extraction results of the two address line extraction units are shown.

When the image shown in FIG. 3A is processed by the miscellaneous information detection unit and the format assumption unit, the format assumption unit has two types of "horizontal and horizontal writing" and "vertical and vertical writing". The format candidates are output, and the two address line extraction units are activated.

Then, the address line extraction unit, to which the format candidate "horizontal writing" is given, extracts the address lines in the form as schematically shown in FIG. 3B, and the width of each address line, Outputs information about length, position, etc. In addition, the address line extraction unit, to which the format candidate “vertical vertical writing” is given, extracts the address lines in the form as schematically shown in FIG. 3C, and the width and length of each address line. , Output information about position, etc.

The format determining unit 15 determines the difference between the default reference value of the address line width and the address line width 61 when the format candidate is "horizontal writing", and the reference value and the format candidate are "vertical". The difference in the address line width 63 in the case of "place vertical writing" is calculated. At this time, if there are multiple address lines, the same processing is performed for other address lines, the cost value of the calculated difference is calculated for each format candidate, and the format with the smaller cost value is adopted. Then, the output of the address line extraction unit using the adopted format is the final output.

When determining the final format, the comparison with the reference value may be performed for the lengths 62 and 64, or for the position and the like. In addition, the width, length, and frequency distribution of the position of the address line are measured in advance from many mail items,
With reference to this, one format can be selected from the format candidates.

The operation of the address line extraction unit will be described in detail below.

FIG. 4 shows a functional block diagram of each address line extraction unit in the address line extraction device of the first embodiment. As illustrated, the address line extraction unit 14 includes a preprocessing unit 21, a labeling unit 22, a block classification unit 23, a character size estimation unit 24, a character block selection unit 25, a base block selection unit 26, and a line verification unit 27. With.

The pre-processing unit 21 binarizes the address area image, removes isolated points, and generates an address area binary image.
The labeling unit 22 performs 8-connected labeling on the binary image of the address area and outputs information about a block that is a minimum enclosing rectangle of each black pixel connected component.

The block classifying unit 23 performs smoothing processing on the histogram obtained by projecting the block filled with black pixels on the axis near the direction orthogonal to the character string direction of the format candidate among the vertical axis and the horizontal axis. Blocks are classified into groups by generating a resulting histogram, and picking up, for each continuous region, blocks that contribute to the continuous region bounded by the portion where the histogram is "0".

In the block classification unit 23, by setting the size of the smoothing mask to a predetermined character size, the characters such as "Hayashi" and "Le" that include separation are included in the character. Also in (1), the histogram is blunted to prevent one line from being divided into a plurality of groups. The blocks classified by the block classification unit 23 into groups are classified by the character size estimation unit 24.
Is output to the character block selection unit 25.

The character size estimation unit 24 calculates the average value of the sizes of the blocks included in each group and sets it as the standard block size in each group. As the size of the block, for example, the width of the block, the length of the long side, the square root of the area, or the like can be used. In the address line extraction device of the first embodiment, the length of the long side of the block is used. ing. The standard block size calculated by the character size estimation unit 24 is output to the character block selection unit 25 and the base block selection unit 26.

The character size estimation unit 24 excludes blocks whose long side length is not included in the range defined by the predetermined upper and lower limits from the processing target and averages the sizes of the remaining blocks. It is also possible to calculate the value and use this as the standard block size.

The character block selection unit 25 determines whether each block belonging to the group is a single addressed character according to the standard block size obtained by the character size estimation unit 24. When the difference between the block size and the standard block size is within a predetermined value, the character block selection unit 25 determines that the block corresponds to a single character, and bases such a block as a character block. Output to the block selection unit 26.

The base block selection unit 26 selects a block having a size closest to the standard block size from the character blocks belonging to each group as a base block for each group and outputs it to the line growth unit 27.

The row growth unit 27 is the base block selection unit 2
The base blocks from 6 are used as blocks to start the integration when the blocks are integrated and the address line extraction processing is performed, and the character blocks belonging to each group are integrated to become the address line candidates. Generate an integrated block. When the format candidate is vertical writing, the blocks that overlap with the base block on the horizontal axis by a certain value or more are integrated,
In the case of horizontal writing, blocks that overlap with the base block on the vertical axis by a predetermined value or more are integrated. The line growth unit 27 is
The same integration is sequentially performed on the integrated blocks, and the blocks are integrated until no more blocks can be integrated in the group. Then, the generated integrated block is output as an address line candidate (format determination unit 15).
Output to).

Second embodiment

The basic structure of the address line extraction device according to the second embodiment is the same as that of the address line extraction device of the first embodiment, but only the function of the address line extraction unit is different. Therefore, only the address line extraction unit will be described.

The character size estimation unit 13 in the address line extraction unit provided in the address line extraction device of the first embodiment calculates the average value of the sizes of the blocks included in each group, and the standard value for each group is calculated. I used the block size. In the second embodiment, the character size estimation unit 13 first excludes the character blocks belonging to the group from the blocks outside the predetermined upper and lower limit values, obtains the average value and standard deviation of the remaining blocks, and averages them. Only when the difference between the value and the block size is within a predetermined multiple of the standard deviation, the average value is calculated again, and the calculation result is set as the standard block size.

Third embodiment

FIG. 5 shows a functional block diagram of an address line extraction unit provided in the address line extraction device of the third embodiment. As is apparent from the figure, the address line extraction unit 14 of the third embodiment
The basic configuration of the address line extraction unit of FIG.
(See), the block dividing unit 28 is added.

The preprocessing unit 21, the labeling unit 22, the block classification unit 23, the base block selection unit 26, and the line growth unit 2
Since the operation content of 7 is the same as the operation in the first embodiment, only the operation of each of the other parts will be described here.

Character size estimating unit 2 in the first embodiment
4 outputs the standard block size to the character block selection unit 25 and the base block selection unit 26.
The character size estimation unit 24 of the present embodiment outputs the standard block size to the character block selection unit 25 and the base block selection unit 26, and also to the block division unit 28.

Then, the character block selection unit 25 selects a block having a size within a predetermined value as a difference from the standard block size as a character block, outputs it to the base block selection unit 26, and outputs the difference from the standard block size. A block having a size larger than a predetermined value and a size larger than the standard block size is selected as a character contact block and is output to the block dividing unit 28.

The block division unit 28 divides the length of the long side of the character contact block by the standard block size and rounds it up to divide the long side of the block into equal parts, and divides the divided blocks into character blocks. Output to the selection unit 26.

That is, in the address line extraction device of the first or second embodiment, when there is a character contact across lines, the address line is extracted by ignoring the block, but the third embodiment. In the example of the address line extraction device, blocks that have become large due to contact are separated in advance, and the separated blocks are input to the line growth unit. For this reason, the address line extraction device of the third embodiment is better than the address line extraction device of the first or second embodiment, for example, when there is a line gap. The address line can be accurately extracted.

Fourth Embodiment

FIG. 6 shows a functional block diagram of an address line extraction unit provided in the address line extraction device of the fourth embodiment. As is clear from the figure, the address line extraction unit 14 of the fourth embodiment
The basic configuration of the address line extracting unit of the third embodiment (see FIG.
(See), the line verification unit 29 is added.

Of the functional blocks shown, the preprocessing unit 2
1, the operation contents of the labeling unit 22, the block classification unit 23, the character block selection unit 25, the line growth unit 27, and the block division unit 28 are the same as those in the third embodiment, and therefore the description thereof will be omitted.

The character size estimation unit 24 outputs the standard block size to the line block verification unit 19 in addition to the character block selection unit 25, the base block selection unit 26, and the block division unit 28.

The line verifying unit 29 determines the width, height, and distance between the character blocks contributing to the integrated block, which the integrated block output from the line growing unit 27 is defined by a predetermined multiple of the standard block size. When the conditions are satisfied, the integrated block is output as the address line candidate.
When the above conditions are not satisfied, a base block change signal is output to the base block selection unit 26, and the generated integrated block is rejected.

Upon receiving the base block change signal, the base block selection unit 26 has the difference from the standard block size within the predetermined value and the closest to the standard block size among the character blocks that have not been the base blocks so far. A block is selected and this is output to the row growth unit 27 as a new base block.

Base block selection unit 26, line growth unit 2
7. The line verification unit 28 repeats the series of processes as described above, and when all blocks in which the address line is stored as a character block are classified into the address line or the base block cannot be selected, the process is performed. Ends (the processing of the address line extraction unit ends).

As described above, in the address line extraction apparatus of the fourth embodiment, it is determined whether the integrated block output by the line growth unit satisfies the condition as the address line, and the determination condition is not satisfied. In this case, the base block is re-elected and the address line integration is redone.

In the address line extraction apparatus of the fourth embodiment, the integrated block has a width, a height defined by a predetermined number of times the standard block size, and a distance between the character blocks contributing to the integrated block. Although the base block is re-elected based on whether or not each condition is satisfied, the line verification unit 29 determines the address that does not satisfy the predetermined condition from the positional relationship of the generated address lines. It may have a function to discard rows. For example, conditions such as (1) the first line has a certain length, (2) the writing position of the first line is written from within a certain area, and the address lines that do not satisfy each condition are discarded. The row verification unit 29 can also be configured to do so.

Further, a statistical table in which the start and end positions of the address line, the size, and the relative position of the lines are described is created in advance, and the likelihood of the address line extracted by this is calculated. , It is also possible to discard rows that do not reach the specified value.

Fifth embodiment

FIG. 7 is a functional block diagram of the address line extraction unit used in the address line extraction device according to the fifth embodiment. As shown in the figure, the address line extraction device 14 according to the fifth embodiment includes an address line extraction unit (see FIG. 6) of the fourth embodiment and a base block selection unit 26 to a character size estimation unit 24. It has been modified so that a signal for feedback can be sent.

The preprocessing unit 21, the labeling unit 22, the block classification unit 23, the character block selection unit 25, and the line growth unit 27.
The operations of the block dividing unit 28 and the row verifying unit 29 are the same as those in the fourth embodiment, and the description thereof will be omitted.

Upon receiving the base block change signal, the base block selection unit 26 has the difference from the standard block size within the predetermined value and the closest to the standard block size among the character blocks which have not been the base blocks so far. A block is selected and is output to the line growing unit 27 as a base block. At this time, if there is no character block that can be a base block, the standard block size correction signal and the addressed line candidate are still in the group. The remaining block, which is a character block that has not been integrated into, is output to the character size estimation unit 24.

When the character size estimating unit 24 receives the standard block size correction signal and the remaining block, the character size estimating unit 24 in the first embodiment will be used for the remaining block.
The standard block size is calculated by the same procedure as in the process in (1) and output to the character block selection unit 25, the block division unit 28, the base block selection unit 26, and the line verification unit 29, and the address line candidate block is generated for the remaining block. I do.

As described above, in the address line extraction device of this embodiment, when the base block selection unit 26 selects a base block, the remaining blocks that are character blocks that have not yet been integrated into the address line candidates are included in the group. If there is no character block that can be used as a base block even if it exists, correct the standard block size again and re-select the base block.

That is, in the fourth embodiment, when there is no block that can become a base block when the base block is selected again, the base block is not selected. For this reason, when rows of different sizes are mixed in one group due to skew, interlacing of rows, etc., despite the presence of residual blocks,
Although the inconvenience that the base block cannot be selected has occurred, this embodiment does not cause such an inconvenience as described above, so that the address line can be accurately extracted.

Below, referring to FIG. 8 to FIG.
The processing for the input image as shown in (a) is executed by the address line extraction device of the fifth embodiment, and the address format assumption unit 1
3 shows an example in which the format candidate is “vertical vertical writing” and the address existence area is the address existence area 30 of FIG. 8B.
The operation contents of the address line extraction device according to the fifth embodiment will be described more specifically.

The preprocessing section 21 having received the information on the address existence area 30 performs a binarization process on the image of the address corresponding to the position of the address existence area 30 to generate a binary image. The labeling unit 22 labels the inside of the binary image generated by the preprocessing unit 21 and outputs a block which is a minimum enclosing rectangle of connected components, as shown in FIG.

In the block classification unit 23, since the format candidate is "vertical vertical writing", each block is projected onto the horizontal axis and the blocks are grouped at the portion where the projection value is "0". As a result, each block shown in (c) has two groups 31 _A , 3 _A and 3 _B as shown in (d).
It will be classified as 1 _B.

As is clear from (b) and (d), at this stage, each group does not always include only one address line, and a plurality of address lines are included in one group like the group 31 _A. May include lines.

The processing procedure for the group 31 _A, which is the group containing the plurality of address lines, will be described below with reference to FIG.

Group 31 _A as shown in FIG. 9 (a)
On the other hand, when the areas of the overlapping portions of the blocks occupy the ratio of the predetermined value of the block area or more, these blocks are integrated as shown in FIG.

As described above, the character size estimation unit 2
Reference numeral 4 is based on the information about the blocks from the block classification unit 23, only blocks having a width included in a range of a predetermined upper limit value and a lower limit value are taken out, and an average value and a standard deviation of the widths of those blocks are taken out. To calculate. Then, a block in which the difference between the calculated average value and the block width is equal to or smaller than a predetermined number of standard deviations is selected, and the average value of the selected block widths is calculated and output as the standard block size.

That is, as shown in FIG.
For a group including such blocks (which is the same as FIG. 9B), first, with respect to the block group shown in FIG. The standard deviation is calculated, and the block group as shown in (c) is selected using the calculation result. And FIG. 10 (c)
The average value is calculated again for the block group as shown in FIG. 3, and the calculation result is used as the standard block size 33, and the character block selecting means 25 and the block dividing means 1
8 to the base block selection unit 26 and the row verification unit 29.

Returning to FIG. 9, the description of the operation of the address line extraction device of the fifth embodiment will be continued. Character block selection unit 25
Indicates that each block in the target group 31 _A is
It is determined whether or not the ratio of the long side to the standard block size determined by the character size estimation unit 24 satisfies the condition of being included in a range of predetermined upper limit value and lower limit value. . Then, a block that satisfies the condition is selected as a character block and is output to the base block selection unit 26, and a block that does not satisfy the condition is output to the block separation unit 27 as a line contact block.

For example, in the group 31 _A shown in FIG. 9B, the block 35 having a large ratio with the standard block size 33 is, as schematically shown in FIG. 9C,
It will be a line contact block, and the other blocks will be character blocks.

The block separation unit 27 divides the width of the line-to-line contact block by the standard block size and rounds it to determine how many line-to-line contact blocks should be separated.
Perform separation. Therefore, the block separation unit 27 determines to separate the block 35 into two blocks by calculating the ratio of the long side length of the block 35 and the standard block size 33. And FIG. 9 (c)
, The separation is performed at the separation position 36, and the two separated blocks are output to the base block selection unit 27.

After all, the base block selection section 27
The blocks shown in (d) are input, and the base block selection unit 27 selects a block 34 having a size closest to the standard block size 33 from those blocks as a base block.

In this case, since the format candidate is "vertical vertical writing", the line growth unit 29 projects the image on the horizontal axis as shown in (e) (that is, in the direction of arrow 38). Blocks that overlap the base block 34 (when projected) (blocks shaded in the figure) are sequentially integrated. Then, the line growth unit 27
When all the blocks in the group are integrated into a character line,
Alternatively, when the integration cannot be performed, the integration process is stopped and the integration result is output to the row verification unit 29. In this case, therefore, as shown in (f), it exists in the right half of the group 31 _A. The integrated result is output as an integrated block 38 when the respective blocks to be integrated are integrated.

The line verification unit 29 checks the size of the generated integrated block, the arrangement of the blocks participating in the integration, and the like, and if the conditions are met, the format determination unit 1 determines that the address line candidate is the address line candidate.
5 is output. Further, the line verification unit 29 checks whether or not there is a block that does not belong to the addressed line in the focused group, and if there is a remaining block,
In consideration of the presence of address lines of different sizes in the group, the character size of the remaining block is estimated again, the standard block size is corrected by the character size estimation unit 24, and the character block selection unit 25 and the block division are performed. The unit 28, the base block selection unit 26, the row growth unit 27, and the row verification unit 29 are caused to execute the processing regarding the remaining blocks again.

Therefore, when there is a block that does not belong to the addressed line as shown in (f), the line verification unit 29 determines that the remaining block is the block in the integrated block 28. Then, the same processing is performed.
As a result, the row growth unit 27 is notified of the new standard block size 41 and the new base block 39, as schematically shown in (g), and the row growth unit 27 integrates the blocks. Then, a new integrated block 42 is generated.

In the address line extraction device of the fifth embodiment, the above operation is continued until there are no remaining blocks in each group or no new address line is generated even if the standard block size is updated, and thereafter. , The process ends.

As described above, in the address line extracting apparatus according to the first to fifth embodiments, the selection of the base block which is the source for generating the address line is the smallest among the blocks existing in one group. Since the configuration is performed based on the average block size calculated without taking account of the factors, the address line is grown from the wrong base block, and as a result, the wrong address line is extracted. It's hard to be told.

That is, as shown in FIGS. 11 (a) and 11 (b), since the size of the address writing character varies in various ways, the base block is selected by a constant parameter regardless of the character size. If the configuration is such that (a)
In order to enable the extraction of address lines composed of relatively small characters (for example, address lines output by word processing) such as the address writing shown in, make the lower limit of the block width that can be the base block small. I have to keep it.
However, if such setting is made, in a mail piece in which a large character is written as shown in (b), as shown by a frame in the figure, for example, a part of the character is a block or It is regarded as a base block, resulting in incorrect address line extraction.

On the other hand, in the address line extraction apparatus of each embodiment, in both cases of FIGS. 11A and 11B, exactly one character block is selected as the base block. The erroneous address line extraction as described above is not performed, and the address line is stably grown.

Further, as shown in FIG. 12A, even in a single address writing, there are many cases in which the size of the character changes line by line such as address name, city / ward name, and street address. However, in the address line extraction device of each embodiment, the image data of such a mail is roughly divided into groups such as groups 31 ₁ and 31 ₂ as shown in FIG. , Address lines are grown in groups.
Therefore, even if the size of characters in one address writing is changed, the address line can be stably extracted.

[0091]

As described in detail above, claim 1 is as follows.
In the address line extraction device according to the invention described, the base block that is the source for growing the address line is selected according to the character size described on the mail, so what character size is used? Even in this case, the address line can be accurately extracted.

In the address line extraction device according to the second aspect of the present invention, the standard block size is calculated without taking into account the character block size distribution and without using an unsuitable character block. Since the base block is selected based on the standard block size, high-precision character size estimation is performed without being affected by noise and character contact, and as a result, accurate address line extraction is possible. You can do it.

Further, as in the invention described in claim 3, when there is a block larger than the standard block size, the address line extraction device is configured to divide the block to generate a character block. In this case, even if the address lines are complicated, it is possible to obtain the address line extraction device that can accurately extract the address lines.

Further, as in the invention described in claim 4, the block integrating means for integrating the blocks (in charge of generating the address line) determines whether or not the integrated result is in the correct form as the address line. By adding a function, it is possible to obtain an address line extracting device with a very low probability that wrong address line extraction is performed.

When the address line extraction device is configured as in the fifth aspect of the present invention, the address line extracting device is configured to detect the address line even when the address lines having different character sizes are interdigitated and contact with each other. The extraction can be done accurately.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing an outline of an address line extraction device according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram for explaining a format assumed by a format assumption unit in the address line extraction device according to the first embodiment.

FIG. 3 shows an outline of a correspondence relationship between an image to be processed and an output of an address line extraction unit, which is used for explaining an operation of a format determination unit in the address line extraction device according to the first embodiment. It is a schematic diagram.

FIG. 4 is a functional block diagram of an address line extraction unit provided in the address line extraction device according to the first embodiment of the present invention.

FIG. 5 is a functional block diagram of an address line extraction unit provided in an address line extraction device according to a third embodiment of the present invention.

FIG. 6 is a functional block diagram of an address line extraction unit provided in an address line extraction device according to a fourth embodiment of the present invention.

FIG. 7 is a functional block diagram of an address line extraction unit provided in an address line extraction device according to a fifth embodiment of the present invention.

FIG. 8 is an image used for explaining the operation of the address line extraction apparatus according to the fifth embodiment, the image to be processed, the information output by the format assumption unit, and the block classification unit in the address line extraction unit output. It is a schematic diagram which showed the outline of the correspondence with information.

FIG. 9 is an explanatory diagram showing an input example of a format determination unit in the present invention for explaining the operation of an address line extraction unit according to a fifth example.

FIG. 10 is an explanatory diagram showing blocks that a character size estimation unit according to a fifth embodiment refers to when calculating a standard block size.

FIG. 11 is an explanatory diagram showing an example of an address line having the same content expressed by characters of different sizes, which is used for comparing the conventional address line extraction technique and the address line extraction device of the embodiment. Is.

FIG. 12 is an explanatory diagram showing an example of an address in which the size of a character changes line by line and how the address is divided into groups.

[Explanation of symbols]

 11 Image Input Section 12 Miscellaneous Information Detection Section 13 Address Format Assumption Section 14 Address Line Extraction Section 15 Format Determination Section 21 Preprocessing Section 22 Labeling Section 23 Block Classification Section 24 Character Size Estimation Section 25 Character Block Selection Section 26 Base Block Selection Section 27 Line growth unit 28 Block division unit 29 Line verification unit

Claims (5)

[Claims] 1. A miscellaneous information detecting means for detecting information other than an address from image data corresponding to the content of the mail, and a detection result of the miscellaneous information detecting means for use in describing the address of the mail. For each of the format candidates assumed by the address format assumption means, which assumes at least one format candidate including the information that specifies the writing direction of the address line,
Block specifying means for specifying a rectangular block having a minimum size including a closed area in which black pixels are connected from the image data for one area corresponding to the format candidate and outputting information on the specified block; A block classification unit that classifies each block into several groups for each format candidate based on the information about the position and size of each block that the block specification unit has identified for each format candidate, and this block classification unit For each group, the average value of the sizes of the blocks belonging to each group is calculated, and the standard block size calculation unit that outputs the calculation result as the standard block size, and the blocks that belong to each group classified by the block classification unit, Standard block calculated for the group by the standard block size calculation means A block whose size difference from the size is within a predetermined value is specified as an address character block, and the specified result is output for each group, and an address character block specification unit is specified for each group by this address character block specification unit. Base block selecting means for selecting, as a base block, an address character block having a size within a predetermined range and having a size closest to the standard block size for the group from the address character block. By the base block selected for each group by, and the address character blocks in the same group having a degree of overlap in the direction orthogonal to the writing direction in the format candidate corresponding to the group is a predetermined value or more, each Output the integrated result for each group as an integrated block. And the block integrating means compares the positions and sizes of the integrated blocks output for each group in the image data with each other in the format candidate unit, and has the most suitable form as an address line. An address line extraction device, comprising: a format determining unit that sets a format candidate for which an integrated block is obtained as a format used for the mail. 2. The standard block size calculation means excludes blocks having a size outside a predetermined upper limit value and a lower limit value from the character blocks belonging to the group, and the remaining block size in the group. The average value of the size and the standard deviation is calculated, and the average value of the sizes of the blocks within a predetermined number of times the calculated standard value and the calculated standard deviation is calculated again, and the calculated result is the standard block. The address line extracting device according to claim 1, wherein the address line extracting device outputs the size as a size. 3. The size of a block belonging to each group classified by the block classifying unit and the standard block size calculated for the group by the standard block size calculating unit by the addressed character block specifying unit. If a block with a difference of less than or equal to a predetermined value is specified as an address character block, and there is a block whose size is larger than the standard block size by a predetermined value or more, then that block is approximately the same as the standard block size. The address line extracting device according to claim 1 or 2, wherein the address line extracting device is configured to divide the block into blocks having a size of, and specify the divided block as an address character block. 4. The condition that the block integrating means includes the integrated block obtained as a result of integration in a range of an upper limit value and a lower limit value of a width or height defined by a predetermined multiple of the standard block size. If it is determined that the condition is not satisfied, the integrated block is discarded, and the base block selecting means uses it as a base block from the character blocks belonging to the group. 4. The address line extraction device according to claim 3, wherein a block which is not processed and whose difference from the standard block size is within a predetermined value and has the smallest difference is selected as a base block. 5. When the base block selecting means does not have a block that can be a base block in a certain group, the standard block size calculating means uses a block which is not integrated in an address line in the group. 5. The address line extraction device according to claim 4, wherein the standard block size is calculated.