JPH03210688A - Line detecting device - Google Patents

Abstract

PURPOSE:To make processing simple, correct, and high-speed, and to make a device small-scale by repeating the generation of marginal distribution, and detecting a character line based on black block position sequence obtained from the marginal distribution. CONSTITUTION:A device is constituted of a marginal distribution generating part 12, a line candidate detecting part 14, and a line detecting part 16. In this case, the starting end and the terminal end in a row direction of a line candidate area are determined by a first marginal distribution obtained by scanning the prescribed processing area of document picture data in a line direction, and the starting end and the terminal end in a line direction of the line candidate area are determined on the basis of a second marginal distribution obtained by scanning the line candidate area in the row direction, and the presence of a divided element in the line candidate area is decided. When there is the divided element, a divided candidate position in the line direction of the line candidate area is set, and the line candidate area is partitioned at the divided candidate position, and division information to show the feature of the character line in a partitioned area is extracted by third marginal distribution obtained by scanning the partitioned area in the line direction, and the position of a line area is determined according to the division information. Thus, the processing can be made simple, correct, and high-speed, and the device can be made small-scale.

Description

[Detailed description of the invention] (Industrial application field) The present invention relates to a row detection device.

(Prior Art) Various studies and studies have been conducted regarding image processing of business cards.

For example, in the conventional technology disclosed in the 37th National Conference of the Information Processing Society of Japan (late 1986), 6W-4p, 1652-1653, the business card image is scanned in the longitudinal direction of the business card to detect the line area. is scanned in the column direction perpendicular to the longitudinal direction to detect black blocks, and the black blocks are scanned in the longitudinal direction and column direction to detect small blocks. And characteristic items on the business card to be processed, for example - Are the multiple small letters that make up the letters placed within a square frame? ! Based on the fact that the characters of items with the same attribute have approximately the same size, etc., small blocks that are considered to constitute a character line of the same item are integrated, and character lines are detected for each item.

Also, Document II: Journal of the Institute of Electronics and Communication Engineers 81/11
Vol, J64-D No, 11 p, 997~
In the prior art disclosed in 10o4, a circumscribed rectangle of a connected region of black pixels extracted from a binary image of a business card is obtained,
After the connecting chains vt of the overlapping circumscribed rectangles are merged (integrated) with each other, the connected regions of the mark portions are separated with reference to the size and position H1r of the circumscribed rectangles. Then, each character line is detected from the peripheral distribution obtained by projecting the connected region excluding the mark portion in the character line direction. Furthermore, by referring to the position and spacing of the circumscribed rectangle, separate the connected area of the title from the character line that is likely to contain the first and last name, and for each character line, divide it into a square frame whose sides are the long sides of the circumscribed rectangle. Integrate the connected areas of the included circumscribed rectangles into one character. Then, the character lines for each item that overlap in the character line direction are separated by referring to the relative value MIW1 with the first and last name, the size of the characters, the number of characters in the character line, etc.

(Problem to be solved by the invention) However, in the prior art of the above-mentioned documents (■) and (H), in order to detect small blocks and connecting chains tfi constituting the - character, and further detect character lines that can be regarded as the same item, small blocks, small blocks, The process is complicated because it is a bottom-up process that repeats the integration and re-division of connected regions, which makes it difficult to realize a small-scale device that can perform line detection at high speed.

SUMMARY OF THE INVENTION In order to solve the above-mentioned conventional problems, it is an object of the present invention to provide a line detection unit M that repeatedly creates a marginal distribution to detect a character line area, instead of repeating the integration and division of small blocks. It's about doing.

(Means for Solving the Problems) In order to achieve this object, the line detection device of the present invention includes a marginal distribution creation unit that scans document image data in a designated area in a designated direction to create a peripheral distribution; a row candidate detection unit that determines the position of the row candidate area in the column direction based on a first marginal distribution obtained according to first instruction information in which the specified area is a predetermined row detection target area and the specified direction is the row direction; and determining the position of the row candidate region in the row direction based on the second marginal distribution obtained according to the second instruction information in which the designated region is the row candidate region and the designated direction is the column direction. Determine whether there is a dividing element in the row candidate area, and if there is a dividing element, set a division candidate value M in the row direction of the row candidate area, set the designated area as a divided area divided by the dividing element of the row candidate area, and specify Extract division information from the third marginal distribution obtained according to the third instruction information whose direction is the row direction, determine the position 11 of the row area according to the division information, and if there is no division element, select a row candidate. The method is characterized by comprising a row detection unit that determines the position of the area as the position of the row area.

(Function) According to the line detection unit M1 having such a configuration, the line detection unit M1 detects the line based on the peripheral distribution of straw obtained by scanning the line detection target area of document image data in the line direction (direction along the character lines). Determine the starting and ending positions of the candidate area in the column direction.

Then, based on the second marginal distribution obtained by scanning the row candidate area in the column direction orthogonal to the row direction, the starting and ending positions of the row candidate area in the row direction are determined, and the dividing elements in the row candidate area are determined. The zero division element whose existence is determined is an element (for example, a blank space or a ruled line) that divides a character line. Therefore, if a division element exists, the line candidate area includes a plurality of character lines.

If there is no dividing element, the position of the line candidate area is detected as the position of the line area.

If a dividing element exists, a dividing candidate value M in the row direction of the row candidate area is set. And the starting end of the row candidate area,
The division IN area divided by the end and division candidate positions is scanned in the row direction, division information is extracted from the third peripheral distribution obtained thereby, and the position of the row area is determined according to this division information.

The division information includes information representing the characteristics of the character lines within the segmented area (for example, the number of character lines, the size and position of the characters constituting the character line, etc.). Characteristics of character lines can be expressed, for example, by the number, width, position, etc. of black blocks within a segmented area. Compare the characteristics of corresponding character lines within the 0 segmented area, and if the features can be considered to be the same, these Character lines are considered to be one character line, and if they are considered to have different characteristics, these character lines are detected as different character lines.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. It should be noted that the drawings are only schematic representations to the extent that the invention can be understood, and therefore, the invention is not limited to the illustrated examples.

FIG. 1 is a functional block diagram for explaining the configuration of an embodiment of the present invention.

As shown in the figure, the row detection device 10 of this embodiment is
A peripheral distribution creation unit 12 that scans document image data in a specified area in a specified direction to create a peripheral distribution, and according to first instruction information that sets the specified area as a predetermined line detection target area and that the specified direction is the line direction. A row candidate detection unit 14 that determines the position of the row candidate area in the column direction based on the obtained first marginal distribution.
Then, the position of the row candidate region in the row direction is determined based on the second marginal distribution obtained according to the second instruction information in which the designated region is the row candidate region and the designated direction is the column direction. Determine whether or not there is a dividing element within, and if there is a dividing element, set the dividing candidate position in the row direction of the row candidate area, set the designated area as a divided area divided by the dividing element of the row candidate area, and specify The division information is extracted from the third marginal distribution obtained according to the third instruction information whose direction is the row direction, and the position 'Fr of the row area is determined according to this division information, and if there is no division element, The line detecting section 16 determines the position of the line candidate area as the position of the line area.

The line position detection device M10 detects the line value N on a document image database such as a business card or a form, and outputs it to the next stage external device, such as a character cutting device.The document image data is generated by the image input unit 18 and is converted into image data. It is stored in the storage unit 20.

This example will be explained in more detail below.

(Image input unit, image data storage unit) The image input unit 18 converts the optical signal from the line detection target area of the document into an electrical signal that is divided into pixels, and outputs document image data consisting of this electrical signal. The unit 20 scans and stores the document image data from the image input unit 18.6 The document image data is, for example, binary image data in which the text portion is expressed with black bits and the text background portion is expressed with white bits.

FIG. 2 is a diagram showing an example of sentence N image data.

The document image data 22 shown in the figure is an example of a business card, in which the 0 line detection target area is set to be slightly wider than the width of the business card to be processed, and the text of the line detection target area is stored in the memory of the image data storage unit 2o. 1 image data is stored. A portion of the image disk 22 that is wider than the line detection target area of the document is expressed with white bits.

An x-y coordinate system is set on the memory, and the 0-line detection target area, which allows the image data of pixel position 1 represented by this coordinate system to be read freely, is defined as O≦X≦xE. And the X-axis direction is defined as 18 character lines (line direction), with 0≦y≦YE.

(Marginal distribution creation unit) Marginal distribution creation] 2 inputs the first instruction information from the row candidate detection unit 14 or the second or third instruction information from the line detection unit 16. Then, based on the specified area and specified direction specified by the instruction information, the document image data in the specified area is scanned in the specified direction, the number of black pixels on the scanning line is accumulated for each sub-scanning position, and the surrounding area related to the specified area is scanned. The zero peripheral distribution for creating the distribution is made up of the cumulative number of black pixels for each sub-scanning position, and the peripheral distribution creation section 12 stores the created peripheral distribution in a peripheral distribution storage section (not shown).

(Line Candidate Detection Unit) The line candidate detection unit 14 of this embodiment sends first instruction information in which the specified area is a line detection target area that includes the entire document and the specified direction is the line direction to the peripheral distribution creation unit 12. Output for.

The line candidate detection unit 14 uses the information indicating the specified area as the scanning start position in the X-axis direction! XS (in this example, XS =
O), scanning end position Xε, and scanning starting position x in the y-axis direction
Ys (YS=O in this embodiment) and the scan end position YE, and also outputs the X-axis direction as information indicating the designated direction.

The marginal distribution creation unit 12 creates a first marginal distribution based on the first instruction information. An example of the first marginal distribution is shown in Figure 3. The vertical axis in FIG. 3 corresponds to the y-axis in FIG. 2, and the horizontal axis represents the cumulative number of black pixels LX (y) on the scanning line at the sub-scanning position 11y. X (y) can be expressed as in the following equation (7).

However, ○≦y≦Y6, and P (x, y) is the coordinate (x,
If the image data at the pixel position of y) is a white bit, P (
If x, y)=O and black bit, then P(x, y)=1.

When the marginal distribution creation unit 12 finishes creating the marginal distribution of sweets, the row candidate detection unit 14 refers to the first marginal distribution,
An area (black block) in which the cumulative number of black pixels equal to or greater than a predetermined threshold LTHL is continuous is regarded as a row candidate area, and the start and end values M in the column direction of each row candidate area are detected. In Fig. 3, the starting end value II in the column direction of the n-th row candidate area counting from the origin ○ side in ascending order of the X coordinate is indicated by the symbol PLT (n), and the terminal value M is indicated by the symbol PL8 (n). (however, n=1.2
,-).

In Figure 4, the row candidate area detected by the row candidate detection unit 14 is shown by a dotted rectangle with the symbol K (n). As shown in the figure, the row candidate area K detected from the first marginal distribution
The starting and ending points N in the row direction of (n) are temporarily set as coordinates X and Xε.

(Line Detection Unit) The line detection unit 16 of this embodiment includes a division candidate position determination unit 22,
It consists of a division determining section 24 and a line position determining section 26.

Main division candidate position determination unit When the row candidate detection unit 14 completes the detection of (n) in the row candidate area, the division candidate position determination unit 22 uses a second As instruction information, the position of the row candidate area K (n) in the X-axis direction WXs, Xt, and the position in the y-axis direction! [Outputs PLT(n), PLB(n), and also outputs y as second instruction information that instructs the specified direction.
Outputs the axial direction.

The peripheral distribution creation unit 12 scans the document image data in the row candidate area K (n) in the y-axis direction based on the second instruction information to create a second peripheral distribution.

5 and 6 show an example of document image data in the line candidate area K (n) and the second peripheral distribution of this area K (n). The horizontal axis in FIG. The coordinate axis corresponding to the X axis and the vertical axis indicate the cumulative number of black pixels LKY(x) on the scanning line at the sub-scanning position x.

LK"l' (X) can be expressed as in the following equation (2).

However, 0≦X≦×6.

When the creation of the second peripheral distribution by the peripheral distribution creation unit 12 is completed, the division candidate position determination unit 22 refers to the second peripheral distribution and divides the continuous area (black block ) is detected, and the starting and ending positions of each black line in the row direction are detected. In Fig. 6, the starting and ending values a in the row direction of the m-th black block counting from the origin ○ side in descending order of the X coordinate are denoted by BXL (m) and BX.
It was expressed as R(m) (where m:=1.2, -).

Then, the row candidate detection unit 14 detects the minimum starting end position (BXL in the example of FIG. 5) of the black block in the row candidate area K (n) in the row direction.
(1)) - is the starting end value 1 in the row direction of the row candidate area K (n)
1KXL (n) and determine the maximum terminal value ml in the row direction (BXR(5) in the example of FIG. 5) as a row candidate l! Area K
The terminal value in the row direction of (n) is determined as 11KXR(n).

Along with this, the row candidate area K is used to determine the presence or absence of a dividing element.
In (n), the WI region (
The width of the white block (hereinafter referred to as a white block) in the X-axis direction is calculated. In Fig. 6, the width of the white block that is # in contact with the end position of the m-th black block in the row direction is indicated by the symbol BBW(i) (however, i = 7.2, -.), white block width. BBW(
i) can be expressed by the following equation (3).

BBW(i)=BXL(m+1)-BXR(m)+1
...-(3) However, if the total number of black blocks detected in the row candidate area K (n) is M, then m is a natural number satisfying 1≦m≦M-1.

Then, the division candidate position determination unit 22 selects one row candidate area K.
The average value width ABBW and maximum value width MXBBW are detected for all white block widths in (n).

The average value width ABBW can be expressed by the following equation (4), assuming that the total number of white blocks in one row candidate area K(n) is l.

Engineering AB8W= (1/I) ・Σ BBW(i)i
=1 ...... (4) Then, a white block having a maximum width MX88W that satisfies the following formula (5) is regarded as a dividing element, and the starting position of this dividing element in the row direction? IIMXL and end position MX are set as division candidate positions DPVI and DVP2, respectively, and information P indicating that the division element exists within the row candidate area K (n) is output. Figure 5 IL In the example shown, DVP
1 = BXR (1) and DVP2 = 8XL (2). In this embodiment, only the white block with the maximum width MXBBW is regarded as a dividing element, and the row candidate area K (n) 1
Consider dividing Fr into two regions.

Also, the following formula (5)! If there is no white block with a satisfying maximum width MX 8 BW, it is assumed that no dividing element exists within the row candidate area K(n), and the determined position KXL ( n), KXRCn>
And the determined position of the column direction? IIPLT (n), PLB
(n) and the division element is the row candidate area K.
Outputs information Q indicating that it does not exist in (n).

MXB8W>ABBW*α1 ・・・・−(5) However, α1 is a predetermined value that is arbitrarily set, for example, α
The ratio is 1:3.

The main division determination partial division determination unit 24 sends information P or Q as to whether or not a division element exists in the row candidate area K (n) to the division candidate position determining unit 22.
The division determination process is performed according to this information P or Q.

When the division determination unit 24 inputs information Q indicating that a division element does not exist in the row candidate area K (n), the division determination unit 24 inputs information q indicating that the row candidate area K (n) is not divided as a division determination result.
Output.

Further, when the division determination unit 24 inputs the division candidate positions 11fDVP1 and DVP2 together with the information P that a division element exists, the division determination unit 24 selects the division area divided by the division element of the row candidate area K (n) according to the third instruction information. Line candidate area K as specified area
(n) column direction 1fPLT(n), PLB(n
), the row direction positions KXL(n), KXR(n), and the division candidate positions 11DVPI and DVP2, and roughly output the column direction as the designated direction of the third instruction information.

The peripheral distribution creation unit 12 generates a region R1 and division candidates between the starting point 11KXL (n) and the division candidate position DVP1 after the row direction of the row candidate region tlK(n) is determined based on the third instruction information. The area R2 from position 1DVP2 to the determined terminal position 1fKXR(n) in the row direction is
A third peripheral distribution LKXL consisting of the cumulative number of black pixels on the scanning line obtained at each sub-scanning position 11y by scanning in the row direction.
(V) and LKXR(y)! create. These third marginal distributions can be expressed by the following equations (6) and (7).

However, in equations (6) and (7), PLT (n)≦y
≦PLB(n).

FIGS. 7 and 8 show examples of the third marginal distributions LKXL(y) and KXR(y).

When the creation of the third peripheral distribution by the peripheral distribution creation unit 12 is completed, the division determination unit 24 checks the third peripheral distribution 89 regarding the segmented region R1, and determines the area in which the cumulative number of black pixels LKXLCV) equal to or greater than the threshold value YLTHL is continuous. (black blocks), and also detect the starting and ending values M of these black blocks in the column direction.

Here, the starting and ending values m+ in the column direction of the first black block counted in ascending order of the y coordinate from the origin side of the segmented area R1 are expressed as BLT (j) and BLB (j), respectively.

Similarly, the division determination unit 24 refers to the third peripheral distribution between the segmented areas R2 and calculates the cumulative number of black pixels LK equal to or greater than the threshold value YRTHL.
Areas (black blocks) where X days (y) are consecutive are detected, and the starting and ending values 11 in the column direction of these black blocks are detected. The starting and ending positions II in the column direction of the first black block counted from the origin side of the segmented area R2 in ascending order of the y-coordinate are expressed as BRT (k) and BRB (k), respectively.

As a result, 11 black block sequences BLT (j), BLB (j) and BRT regarding segmented regions R1 and R2 are obtained.
(k), BRB (k)8 is obtained.

The division determination unit 24 focuses on the adjacent segmented regions 1 and R2 in the row candidate region K (n), and uses the black block position series to determine division information (for example, the width of the black block, etc.) between these regions. A center position N) is created and a division judgment is made according to the judgment criteria ① to ② below. If one of the judgment criteria ■～■ is satisfied, information p is output that indicates that the character lines of circle 1 and R2 in the focused segment area have different character lines from each other, and the judgment criteria ■～■ If all of the conditions are not satisfied, information q indicating that the character lines in the target segmented regions R1 and R2 are considered to have the same characteristics is output.

■ The total number of black blocks in the first two segmented areas is not equal.

■If the total number of black blocks in the two divided areas of interest is the same, the numbers counted from the origin side of both areas in ascending order of the y coordinate are the same, and the difference in width in the column direction of the black blocks is greater than or equal to a predetermined value. Yes, there are at least one set of black blocks like this.

■: When the total number of black blocks in the two divided areas of interest is the same, the numbers counted in descending order of the y-coordinate from the origin side of both areas are the same, and the center position of the black blocks in the column direction is greater than or equal to a predetermined range in the column direction. By the way, there are at least one pair of black professional Tsuku like this.

Hereinafter, with reference to FIG. 9, the flow of the division determination process will be explained by giving an example. FIG. 9 is a diagram showing the flow of division determination processing. In the following description, the total number of black blocks in the divided regions R1 and R2 is assumed to be nine and two.

When the marginal distribution creation unit 12 finishes creating the third marginal distribution, the division determination unit 24 immediately starts processing), and then selects the focused segment area t! It is determined whether the total numbers J and K of black blocks in 1iR1 and day 2 are equal, and the criterion (2) is checked (Sl).

If the total numbers J and K are not equal, it is assumed that the character lines in the target segmented region R1 and the character lines in Day 2 have different characteristics, and the line candidate region K (n) is divided into the focused segmented region Day 1, R
Please output the information p indicating that it is divided into two.S13)
, and then ends the process (S13).

If the total number J, is equal, the control variable β has an initial value =
After setting the value to 0, first, in steps 33 to S6, the criterion (2) is checked. Since the total numbers J and K are equal, K will be represented by J in the subsequent processing.

In 83 to S6, 1 is added to the control variable β (R3), and then the position of the first black block of the focused segmentation area day 1 and day 2 BLT l), BLB (β) and BRT (β),
From BRB(β), find the black block widths BLW and BRW in the Mβ-th column direction t in area 81 and day 2 (
R4), these black block widths are expressed by the following equations (8) and (9).
It can be expressed as

BLW=BLB 1)-BLT (42)+1-(8
) 8RW=8R8(β)-BRT(A)+1-(9) Next, the difference between these black block widths BLW and BRW is set to a predetermined value β.
, (R5), and in this example, the difference in black block width at R5 is IBLW-BRWI7i! I tried to compare it with the absolute value β, but the difference IBLW-8R
WIIFr is a relative value βI *Min (BLW
, BRW).

BLW-8RWl>8+ *Ml n (8LW, B
RW) ・・・・・・(1
0) However, Min (8LW, BRW) is BLW, SR
Represents the smaller value of W.

When the difference in black block width exceeds the predetermined value β, the process of S13 is performed, and when the difference is less than the predetermined value 81, it is determined whether the number β is equal to the total number J (B6).

If the number β is 1 equal to the total number J, then the other group's black professional
After that, the process returns to step 83 to check the criterion (2), and if the number l is equal to the total number J, then the processes 87 to S11 are performed to check the criterion (2).

In steps 87 to S11, the control variable 1 is first set to an initial value of 0 (S7), and then 1 is added to the control variable β (B8).

Next, the first black block digits 11BLT (β), 81B (I2) and BRT of the attention division wI area day 1 and B2
(β), 'BRB (Il), black block center position 11BLC opened in the direction of the β-th column of regions R1 and R2
and BRCt (B9), c: These center positions can be expressed by the following equations (11) and (12).

BLG= (BLB (β)+BLT l))/2−(
11) BRW= (BRB (β) +BRT (,1ri)
/2...(12) Next, the difference between these center positions 8LC and BRC is compared with a predetermined value B2 (S10). In this example, the difference in black block width I is 810. value β
I tried to compare it with 2, but there is a difference + 8LC
−BRClj & For example, a relative value depending on the center position as expressed by the following equation (13) 82 *Min (BL
C, BRC).

I BLC-8th CI>B2 *Min (BLCl
BRC) --------(1
3) When the difference in center position exceeds the predetermined value β2'Fr, S13
and when the difference between them is less than a predetermined value B2, it is determined whether the number β is equal to the total number J (31
1).

If the number .beta. is not equal to the total number J, the process returns to B8 to check the criterion (2) for other sets of black blocks, and if the number .beta. is equal to the total number J, the process is terminated (end).

Vehicle position determination section The row position determination section 26 receives information p or q from the division determination section 24.
The row value Wl is determined accordingly.

In the two divided regions R1 and Day 2 that we focused on here,
The third black block detected from the third peripheral distribution of section Il area R1 and counted in descending order of y coordinate is designated by code 81(j
), the starting and ending positions of the black block 81 (j) in the column direction are BLT (j) and BLB (j), and the starting and ending positions of the black block 81 (j) in the row direction are classified as The starting value @KXL (n) and the ending value @DVPI of the region R1 are obtained. Note that ]=1.2, -J.

Similarly, y detected from the third peripheral distribution of segmented region R2
Counting the coordinates from smallest to smallest, the first black block is coded 8.
2(j), the starting and ending positions of the black block B2(j) in the column direction are BRT(j) and BRB(]), and the starting and ending positions of the black block B2(j) in the row direction are BRT(j) and BRB(]). The positions are the starting edge value 11DVP2 and the ending edge position KXR(n) of the segmented area R2, where k=1.2, . . . , K.

Information p! In the case of a large input, the total number of black blocks in each of the segmented regions R1 and R2 is equal, so J=, so with j=, the row position determination unit 26 selects the IEj-th black block 81 (j), 82(j) as one character line Zp(j). In this case, the character line area Zp (
The starting and ending positions of black block B1(j) in the row direction are the starting end value WKXL (n) of the black block B1(j) and the ending value 11KXR(n) of the black block 82(j) in the row direction.
And let the starting end position of the character line area Zp (j) in the column direction be the minimum starting end position of the starting end values 11BLT (j) and BRT (j) of the black blocks 81 (j) and 82 (j) in the column direction. and the terminal value i in the column direction of this W4 area Zp(k)! If! This is the maximum end position among the end positions BLB(j) and 8RB(j) of the black blocks B1(j) and 82(j) in the column direction.

In addition, when information q is input, the Mj-th black block 81(j) in one segmented area R1 is divided into one character line 91.
Let t*Za1(j) be the starting end position and end value M in the row direction of this area Zql(j) as the starting end position KXL(n) and end value @DVP1 in the row direction of the black prong 81 (j),
In addition, the starting and ending positions of this area ZQ1(j) in the column direction are the starting end positions of the black block Bl(j) in the column direction BLT(
j) and the terminal value WBLB(j).Similarly, the second black block B2(k) of the other segmented area 2 is defined as one character line area ZQ2(k), and this area Zq2
Starting and ending values Wl in the row direction of (k) & black block 8
2 (k) in the row direction starting end value 1fDVP2 and ending position KXR(n), and this 1ItfiZc+2(k)
The starting and ending values N in the column direction of the black block 82 (k)
The starting end value 1fBRT (k) and the ending value 11B in the column direction of
Let it be RB (k). In addition, if you enter information q, enter J
=K or J≠.

Figures 0 to 13 show examples of detection results of character line areas according to this embodiment. Figure 10 is image data 22 of Figure 2
The detection result of the character line area between
Character line area ZQ detected from the line candidate area that does not satisfy
1 (j) and Zq2(k), Figure 12 shows the character line area Z detected from the line candidate area that does not satisfy the criterion ■.
Examples of QI(j) and ZQ2(k), and FIG. 13 are diagrams showing examples of character line areas Zql(j) and ZQ2(k) detected from line candidate areas that do not satisfy the criterion ■. .

In these figures, the detected character line area is indicated by a rectangular frame with two-dot chain lines, and the character line area Zp when 0% information is input.
An example of the detection result of (j) is expressed with the symbol Zp (]),
Similarly, when information q is input, the character line area ZQ 1 (
j), ZQ2 (k) has the code ZQ 1 in the detection result example.
(j) and ZQ2 (k) are shown as appendixes.

As can be understood from FIG. 10, in the image data 22 of FIG. 2, all row candidate regions other than the row candidate region corresponding to the name writing region are divided according to the above-mentioned criterion (2). In addition, the line candidate area corresponding to the name writing area satisfies the above equation (5) and has division candidate values 11DVP1, DVP
2 is set, but it is not divided because it does not meet any of the above criteria ■ to ■, and the character line that includes both the corresponding black blocks 81 (j) and 82 (j) in the segmented areas tlJR1 and R2. Region Zp (j) will be detected.

The present invention is not limited to the above-described embodiments; therefore, the operating conditions, flow of operation, input/output signals, numerical conditions, and other conditions of each component can be changed as desired.

For example, instead of the above equation (5), the division elements may be detected and the division candidate positions may be set according to the following equation (14).

BBW(i+1)-88W(i)l>02...(14
) However, if the total number of white blocks sandwiched between black blocks detected from the Soji marginal distribution in the row candidate area is 1, then
1 is a natural number such that 1≦i≦k.

In this case, the variation in the width of the white block in the row direction is checked in order of decreasing X coordinate according to equation (14), and the white block width 88W (i+1), which has a variation of more than a predetermined value of 02, is detected as a dividing element. Based on this dividing element, dividing candidate position DVP1 is determined.
, set DVP2.

(Effects of the Invention) As is clear from the above description, according to the line detection device of the present invention, based on the first marginal distribution obtained by scanning a predetermined processing area of document image data in the line direction, Determine the starting and ending positions of the row candidate area in the column direction. Next, determine the starting and ending positions of the row candidate area in the row direction based on the second peripheral distribution obtained by scanning the row candidate'ga stage in the column direction. 8% is determined, and the presence or absence of a dividing element within the row candidate area is also determined.

If there is no dividing element, the position of the line candidate area is detected as the position of the line area.

In addition, if there is a dividing element, a dividing candidate position in the row direction of the row candidate area is set, the row candidate area is roughly divided by the dividing candidate position, and this divided area is scanned in the row direction. Based on the third marginal distribution, division information representing the characteristics of the character lines within the divided area is extracted, and the position of the line area is determined according to this division information. In this position determination, the characteristics of corresponding character lines in the two divided areas of interest are compared, and if the characteristics can be considered to be the same, these character lines are treated as one character line, and if the characteristics are different, then the character lines are treated as one character line. If possible, these character lines are detected as different character lines.

Therefore, according to the apparatus of the present invention, character lines can be detected based on the black block position series obtained from the peripheral distribution by repeatedly creating the peripheral distribution even for a document image with a complicated line arrangement as shown in FIG. 10, for example. For this reason, it is possible to detect small blocks etc. from a document image and integrate or re-divide the small blocks etc. based on the size and position of the circumscribed rectangle of the small blocks etc. and the rules between the document images to be processed. Compared to conventional devices that repeatedly perform line detection from complicated document images, the line device can accurately perform line detection with simpler processing.

Furthermore, since line detection can be performed through simpler processing, the apparatus can be made smaller and the processing speed can be increased.

[Brief explanation of drawings]

Figure 1 is a functional block diagram for explaining the configuration of an embodiment of the present invention, Figure 2 is a diagram showing an example of document image data, and Figure 3 is a first marginal distribution regarding the document image data in Figure 2. Figure 4 is a diagram showing the line candidate area detected from the first peripheral distribution in Figure 3 by the line detection unit, Figure 5 is an image of the line candidate area hidden in the document image data in Figure 2. Figure 6 is a diagram showing the 20th marginal distribution between the image data in Figure 5. Figures 7 and 8 are diagrams showing the third marginal distribution regarding the image data in Figure 5. , Figure 9 is a diagram showing an example of the flow of division determination liquid processing, Figure 10 is a diagram showing the line detection results between the document image data of Figure 2, and Figures 11 to 13 are diagrams showing the line detection results. It is a figure which shows another example. 10... Row detection device, 12-. Marginal distribution creation unit 14
---Line candidate detection unit, 16...Line detection unit 20...Image data storage unit

Claims (1)

[Claims] (1) A peripheral distribution creation unit that scans document image data in a designated area in a designated direction to create a peripheral distribution; a row candidate detection unit that determines the position of the row candidate area in the column direction based on the first marginal distribution obtained according to the instruction information; Based on the second marginal distribution obtained according to the second instruction information, the position of the row candidate area in the row direction is determined, and the presence or absence of a division element in the row candidate area is determined, and if the division element exists, A dividing candidate position in the row direction of the row candidate area is set, the specified area is set as a segmented area divided by dividing elements of the row candidate area, and the designated direction is set as the row direction. extracting division information from the third marginal distribution, determining the position of the row area according to the division information, and determining the position of the row candidate area as the position of the row area if the division element does not exist. 1. A row detection device comprising: a detection section.