JPH03139783A - Method for segmenting circumscribing rectangle of document image - Google Patents

Abstract

PURPOSE:To reduce memory capacity used in processing and to shorten processing time by judging connectivity each time as scanning a binary image with a circumscribing rectangle detecting means, and forming and preserving a circumscribing rectangle sequentially. CONSTITUTION:The method is comprised of a binarization processing means 10 to obtain the binary image by binarizing a multivalue image, and a circumscribing rectangle segmenting means 20 to segment the circumscribing rectangle by inputting the binary image, and the circumscribing rectangle segmenting means 20 is equipped with a mask setting means 21 and the circumscribing rectangle detecting means 22. The mask setting means 21 performs the setting of a mask with arbitrary size by scanning binary image data obtained by binarization. Also, the circumscribing rectangle detecting means 22 generates and preserves the information of the circumscribing rectangle sequentially from, for example, a black picture element and a remarked picture element appearing within the range of the mask generated at the mask setting means 21 as scanning the binary image, therefore, the processing is performed in unit of picture element(unit of line). Thereby, the memory capacity is reduced and the processing time is shortened.

Description

Detailed Description of the Invention (Field of Industrial Application) The present invention is applicable to document processing methods such as optical character reading (OCR), region segmentation, character string extraction, and index extraction. The present invention relates to a method for cutting out a circumscribed rectangle of a document image containing various elements such as characters.

(Prior art) Conventionally, as a technology in this field, there is a document 1: "Image Signal Processing for rFAX and OA" by Fukinuki, first edition (1982).
-10-20') Nikkan Kogyo Shimbun, P, 5-25, Document 2; Japanese Patent Application Laid-Open No. 103372/1983, Document 3; Transactions of the Institute of Electronics, Information and Communication Engineers, J71-D [12] (
1988-12) P, 2587-2595, Reference 4; Karasuwaki [Digital image processing for image understanding (II) J (1988-2> Shokodo Publishing Co., Ltd., P, 44)
-47, etc.

Document 1.2 describes a technique for binarizing a document image to obtain a binary image, and Document 3.4 describes a method for cutting out a circumscribed rectangle of a document image.

In the method described in Reference 3, after storing a binarized (white: background color, black: text button color) digital image (one frame worth) in a memory, main scanning is performed from the upper left corner of the memory. direction (or sub-scanning direction), one black pixel (
In this method, a boundary point) is found, and adjacent boundary points are sequentially detected using that point as a starting point to go around the boundary once. In particular, in the method of Document 3, instead of using the conventional 3×3 mask, the vertical and horizontal dimensions are (2m+1>X (2n+1.>(m).

Using a boundary tracking mask (n is a positive number), we extended the concept of 8-connectivity by changing the values of m and n according to the purpose, and performed boundary tracing of connected components at various levels. Minimum X for each area.

Circumscribed rectangles are extracted using the y coordinate value and the maximum x and y coordinate values.

Furthermore, in the method described in Reference 4, the connectivity of black pixels is checked in the process of scanning in the sub-scanning direction and in the main scanning direction for a digital image that has been binarized in the same manner as described above. Then, each connected component is labeled pixel by pixel (processing to obtain a label image from a binary image), and the result is saved as a label image. And from that label image,
The minimum x, y coordinate values and maximum x of each component (label number),
This was a method of extracting a circumscribed rectangle based on the y-coordinate value.

(Problems to be Solved by the Invention) However, the above method has the following problems.

(a) In the method of Reference 3, when the number of connected components increases, the tracking/extraction process takes a very long time.
Since all images must be stored in memory,
There was a problem that the memory capacity increased.

(b) The method in Reference 4 requires memory to store one frame's worth of label images for processing,
Its memory capacity becomes large. Furthermore, only 4-connectivity and 8-connectivity are referenced, and there is a problem in that it is not possible to extract connected components at various levels by expanding the concept of 8-connectivity.

The present invention provides a method for cutting out a circumscribed rectangle of a document image, which solves the problems of the prior art, such as long processing time, large memory capacity, and the fact that only 4-concatenation and 8-concatenation can be supported. It is.

(Means for Solving the Problems) In order to solve the above-mentioned problems, the first invention provides a method for cutting out a circumscribed rectangle of a document image.
The value image is input to a mask setting means, and after arbitrarily setting a mask indicating a certain area to determine that all black pixels appearing in a certain area are in the same circumscribed rectangular area, the mask setting means The rectangle detection means scans the binary image according to the mask set by the mask setting means, determines connectivity each time, and sequentially forms and stores circumscribed rectangles (rectangles and squares).

In a second aspect of the invention, in the first aspect of the invention, if the portion of unscanned pixels beyond the position of interest that has not yet been scanned is not included within the range of the mask,
The size and shape of the mask can be set arbitrarily.

In a third invention, in the first invention, the circumscribed rectangle detection means determines that all black pixels appearing within the mask range created by the mask settings are the same circumscribed rectangular area, and converts the binary image into While scanning, the circumscribed rectangles are sequentially formed and saved.

In a fourth invention, when the circumscribed rectangle detection means in the first invention is executed, a label (memory location) in which circumscribed rectangle information corresponding to the size of the mask is stored is stored in a label storage line memory, By storing different labels existing within the arbitrarily set mask in a label storage register and further storing the circumscribed rectangle information in a circumscribed rectangle position memory, the circumscribed rectangles are sequentially formed and saved.

In a fifth invention, in the fourth invention, the label storage line memory stores nl of the set mask.

02 minutes and the size of the image are set as one label storage line memory, and the position of the label storage line memory corresponding to the coordinate values (x, y) of the binary image is (x, y) of the predetermined label storage line memory. -1-nl).

In a sixth invention, the label storage line memory and label storage register in the fourth invention store the label stored in the circumscribed rectangle information to which the pixel belongs in the case of a black pixel, and otherwise (in the case of a white pixel) ) stores the initial value (0) as it is, and in the fourth invention, when the number of different label numbers stored in the label storage register is two or more,
a first step of integrating and storing circumscribed rectangle information corresponding to a number of labels (where a is an arbitrary number) stored in the label storage register; and when the number is 1.2 or more; , the second step of integrating the coordinate values (x, y) of the current position of interest and the circumscribed rectangle information corresponding to its label number and saving it at the same label position; A third step of adding information on the coordinate values (x, y) of the position as circumscribed rectangle information is executed.

In an eighth invention, in the seventh invention, the a number of labels stored in the label storage register are sorted in descending order of A1,
Arrange A2, ...Aa, compare the label information of A1 with other label information to find new information, save the result to the label of A1, and delete the information of the integrated label. Then, the information of subsequent labels is moved to the previous label, and the contents of the label storage line memory and label storage register are also rewritten accordingly, so that the maximum label always indicates the number of circumscribed rectangles as it is. ing.

(Function) According to the first invention, since the method for cutting out a circumscribed rectangle of a document image is configured as described above, the mask setting means scans the binary image data that cannot be obtained by binarizing it and generates an arbitrary image. Set the size mask.

While scanning the binary image, the circumscribed rectangle detection means sequentially creates and stores information about the circumscribed rectangle from, for example, black pixels and pixels of interest that appear within the range of the mask created by the mask setting means. This makes it possible to process on a pixel-by-pixel (line-by-line basis), reducing memory capacity and processing time.

Second. In the third invention, circumscribed rectangles under various conditions (levels) can be extracted by arbitrarily setting the size and shape of the mask using the mask setting means.

In the fourth to eighth inventions, by using the label storage line memory, the label storage register, and the circumscribed rectangle position memory, the process for extracting the circumscribed rectangle can be simplified and simplified.

Therefore, the above problem can be solved.

(Embodiment) FIG. 1 shows an embodiment of the present invention, and is a schematic functional block diagram for explaining a method for cutting out a circumscribed rectangle of a document image.

1 This functional block binarizes a multivalued image and converts it into a binary image (
A binarization processing means 10 obtains white (0): background pattern color, black (1): character pattern color), and a circumscribed rectangle cutting means 20 inputs the binary image and cuts out a circumscribed rectangle.
It is configured.

The binarization processing means 10 is a CCD (Charge Co.
The image sensor includes a reading unit 11 that scans an information medium using the image sensor to obtain a multivalued image. On the output side of this reading unit 11, there is a RAM (
A multivalued image memory 12 such as Random Access Memory> and a binarization means 13 for converting the multivalued image in the multivalued image memory 12 into a binary image are connected. Further, on the output side of the binarization means 13, the 2
A binary image memory 24 such as a RAM that stores the binary image converted by the value converting means 13 is connected.

The circumscribed rectangle cutting means 20 cuts out a circumscribed rectangle by scanning the binary image stored in the binary image memory 14, and all black pixels appearing within a certain two-area range of the binary image are cut out from the same circumscribed rectangle. A mask setting means 21 is provided for setting a certain area range for determining that the area is a rectangular area.

Connected to the output side of the mask setting means 21 is a circumscribing rectangle detecting means 22 that sequentially forms and stores circumscribing rectangle information from black pixels and pixels of interest that appear within the set mask range. Furthermore, this circumscribed rectangle detection means 22 includes a label storage line memory 23 such as a RAM that can store the circumscribed rectangle information (memory location and label) to which each black pixel belongs corresponding to the size of the mask. A label storage register 24 corresponding to the size of the mask (actually, it can be quite small) capable of storing different label numbers that appear within the mask range, and a RAM necessary to sequentially form and store circumscribed rectangle information. A circumscribed rectangular position memory 25 such as the above is connected.

Next, the method for binarizing a multivalued image (■) and the methods for cutting out circumscribed rectangles (II) to (I[I) will be explained with reference to FIGS. 3 to 7.

(I> Binarization method for multivalued image The reading unit 11 in FIG. 1 scans the information medium to output a multivalued image, and stores it in the multivalued image memory 12. The finer the reading resolution, the better, 8 lines/
A resolution of mm or more is appropriate. The binarization means 13 reads the multivalued image data in the multivalued image memory 12, converts the multivalued image data into a binary image, and converts the multivalued image data into a binary image.
Store in 4. 2 stored in this binary image memory 14
In the value image data, the character slam pixel is a black pixel (black bit)
, background slam pixels are represented by white pixels (white bits).

These binarization methods can be realized, for example, by the method described in Document 1.2.

(II) Mask setting method The mask setting means 21 scans the binary image data in the binary image memory 14 obtained by the binarizing means 13 to set a mask. Examples of mask settings for scanning direction 1 are shown in FIGS. 2(a) to 2(e), and FIG. 3(a) shows an example of mask settings for scanning direction 2.

Shown in (b).

Although the main scanning direction of the binary image data in the binary image memory 14 may be either the X or Y axis direction, in this embodiment, the main scanning direction of the binary image data in the binary image memory 14 is as shown in FIGS. 2(a) and 3(b). ni, 2
The main scanning direction of the value image data is the X-axis direction, and the sub-scanning direction is the Y-axis direction. Also, the scanning direction is shown in Fig. 2(a).
You may scan from the top left of the image in the direction of the arrow, as in scanning direction 1, or you may scan from the top left of the image in the direction of the arrow, as in scanning direction 2 of FIG. 3(a). .

The mask setting means 21 sets a mask for determining that all black pixels appearing within a certain area are in the same circumscribed rectangular area according to the scanning direction 1.2. Therefore, a mask setting method in the case of scanning direction 1 will be explained using FIGS. 2(b) to 2(e), and a mask setting method in the case of scanning direction 2 will be explained using FIG. 3(b).

As shown in FIGS. 2(b) to 2(e), the mask can be set to any size. For example, the mask shown in FIG. 2(b) corresponds to 4 connections, and the mask shown in FIG. 2(C) corresponds to 8 connections. Further, the example of mask setting shown in FIG. 3(b) corresponds to the same 8-connection as in FIG. 2(C).

In any of these masks, the mask is set around the pixel of interest, but it is set so that the portion of pixels beyond the pixel of interest (that have not been scanned yet) are not included within the range of the mask.

Here, since the next circumscribed rectangle detection means 22 performs processing according to the size and shape of this mask, information necessary for the processing will be explained.

First, the pixel position in the mask as seen from the pixel of interest is L1, L
2...Let Lη be the number of pixels, X of the mask
The maximum number of pixels in the direction is m, and the maximum number of pixels when viewed to the left (in the direction of the small X) from the pixel of interest in the x direction is nl. Similarly, the maximum number of pixels when viewed to the right (the direction of the larger X) from the pixel of interest in the X direction is 02. For example, in the mask of FIG. 2(C), 1 is 4 (L1, L2.L3.L4)
, m is 2, nl is 1, and n2 is 1.

(III) Circumscribed rectangle detection method The circumscribed rectangle detection means 22 determines that all black pixels appearing within the mask are in the same circumscribed rectangle according to the mask created by the mask setting means 261, and detects the circumscribed rectangle while scanning the image. Process to form and store sequentially. Examples (i) to (iii) of this processing procedure are shown in the circumscribed rectangle detection flowchart of FIG. Note that 830 to S48 represent each step.

(i) Step S30 in FIG. S31 First, input the mask information created by the mask setting means 21 (530).
), the label storage line memory 23 and the label storage register 24 are set (S31). The method will be explained using FIG.

For example, as shown in Figure 5, the size of the image is changed to cols
In Xrows, the mask is 11 = 4 (LL,.

L2. L3. L4) , rrr-2, n1=1, n21
When set with an 8-connection mask (see Figure 2 (C)), a label storage line memory that can store cols+nl (=1>+n2(-1) memory locations (labels) of circumscribed rectangle information) is used. m (=2) pieces of 23, and N (
-4) Prepare a label storage register 24 that can store different labels (actually, since the labels of the circumscribed rectangle information to which adjacent black pixels in the mask belong are the same, the number of labels to be stored in the label storage register 24 is ).

Here, information on the current scanning line is stored in the label storage line memory 2B-1, information on the previous line is stored in the label storage line memory 23-2, and so on, information on the mth previous scanning line is stored in the label storage line memory 23-1. It shall be stored in m.

Furthermore, the relationship between these is that when the current position of interest is (x, y) of the image, as shown in FIG.
shall be stored in a location located at This means that when the position of interest is (1, y) in the image, L1 shown in the mask
,, This is to simplify the processing by taking the LJ portion into consideration. Similarly, when the position of interest is (cols, y) of the image, the L3 portion of the mask is also taken into account in the processing. Therefore, a label storage line memory 23 capable of storing cols+nl+n2 items is required. moreover,
1 (-4) L1, L2 in the mask...
・, LJ) indicates the position within the mask,
For example, in the case of FIG. 5, Ll is the line memory 23-2.
(x+n1-1>, L2 is (x+n1-1>, L2 is (x+n1-1>,
x+n1), L3 is (x+n1) of the line memory 23-2
l+1>, LJ is (x+n1) of line memory 23-1
-1. ).

Then, while scanning the image, circumscribed rectangles are sequentially formed,
The information is stored in the circumscribed rectangle position memory 25. In addition, label storage line memory 23 and label storage register 2
4, if a black pixel exists within the mask range set around the current position of interest, the circumscribed rectangle position memory 2
It is assumed that a memory location (label) indicating a storage location of circumscribed rectangle information to which this black pixel belongs in 5 is stored.

After considering the above points, each memory is set in step S3 in FIG. 4.

(ii) Next to steps 332 to 348 in FIG. 4, m label storage line memories (23-19 to 23-m) & two are initialized (S32).

This is for each label storage line memory (cols+n1 tenn
O, which is the label information of white pixels, is assigned to each of the (2) arrays as an initial value. That is, if the initial value is 0, it is a white pixel, and if one or more label information is written, it is a black pixel.

When all the label storage line memories 23-1 to 23m have been initialized, y is then initialized to the initial value 1 (333).
. Then, the information in the label storage line memory 23-1 is transferred to the label storage line memory 23-2, and the information in the label storage line memory 23-2 is transferred to the label storage line memory 23-3.
Shifting (S34) of the line memory is performed to move the information in each memory to the label storage line memory 23-(m-1>) to the label storage line memory 23-(m-1). 23-1 remains the same)
.

Next, X is initialized with an initial value of 1- (S35).

Then, the (X10 n1,)th label storage line memory 23-1 is initialized (S36). This initializes the label storage line memory 23 at the current position of interest, and similarly to 332, O, which is the label information of the white pixel, is substituted into that position as an initial value.

Then, it is necessary to determine whether the point of interest W (x, y) in the image at that time is a black pixel or a white pixel (537), and if it is a black pixel, then
Within a mask range (LL, L2.
・1 in the part corresponding to the position of L, Il
All (a) different label numbers above (other than the initial value) are stored in the label storage register 24 (S38).

Next, if the number a of different label numbers stored in the label storage register 24 is two or more, the number of different label numbers stored in the label storage register 24 is stored in the circumscribed rectangle position memory 25 corresponding to the a label numbers stored in the label storage register 24. The circumscribed rectangle information is 1
and save it (S40). Furthermore, the current position information (x, y) and the information stored in the circumscribed rectangle position memory 25 corresponding to the label number are integrated and stored at the same label position (841), and the stored label number is added to the label. It is saved at the current position (x+nl) of the save line memory 23-1 (843).

After that, it is determined whether it has been performed for all x (cols) 544), and if it has been performed for all X, then it is determined whether it has been performed for all y (rows) 545). Tara,
The process ends (346).

If it is determined in S45 that all y have not been processed yet, the value of y is incremented (548),
Return to S34. If it is determined in S44 that not all X have been processed yet, the value of X is incremented by 54
7), return to S36.

In S39, if the number a of different label numbers stored in the label storage register 24 is 1, then 841 is executed, and if a is 0, the coordinate information of (x, y) is newly added to the circumscribing rectangle. In addition to the position memory 25, steps 542) and 343 are executed. Further, if it is determined in S37 that the pixel of interest (x, y) is white, S44 is executed.

(iii) The operations of steps 840 to 843 and steps 340 to 84B in FIG. 4 will be explained in more detail with reference to FIGS. 6 and 7.

FIG. 6 is a flowchart showing an example of the operation of 840 in FIG.

In FIG. 6, first, a number of labels (A1, A2, etc.) are stored in the label storage register 24.

Aa) is input (S50). After initializing i with the initial value 2 (S51), the circumscribed rectangle information in the circumscribed rectangle memory 25 corresponding to the labels of At and Ai is compared, and the minimum x
, the y coordinate and the maximum x, y coordinates are respectively the starting point of the circumscribed rectangle,
The information is stored as the end point coordinates in the circumscribed rectangle position memory 25 corresponding to the label of A1 (S52). Next, the label Ai in all the label storage line memories 23 is changed to the label A1 (853). Then, the circumscribed rectangle information corresponding to the label of Ai in the circumscribed rectangle position memory 25 is deleted (554), and when i becomes a (S55), the process is terminated (S56).
In step 5, if i is not a, the value of i is incremented (557) and the process returns to step S52.

S54 initializes the label Ai in the circumscribed rectangle position memory 25, and thereby it can be determined whether circumscribed rectangle information has been written or not. In addition,
Circumscribing rectangle information may be saved to be newly added to the label deleted in S54.

FIG. 7 is a flowchart showing another example of the operation of 840 in FIG.

In FIG. 7, first, a number of labels stored in the label storage register 24 are sorted in ascending order of A1, A2, etc.
. , Aa and input them (S60).

After initializing i with the initial value 2 (S61), the circumscribed rectangle information in the circumscribed rectangle position memory 25 corresponding to the labels of Al and Ai is compared, and the minimum x, y coordinates and the maximum x, y coordinates are determined. The starting point of each circumscribed rectangle.

4 Set new end point coordinates and save the information in the circumscribed rectangle position memory 25 corresponding to the label of A1 (S62
). Label AI in all label storage line memories 23
is changed to the label of A1, and those with labels larger than Ai are decremented (863). Furthermore, Ai+1 of the label storage register 24. Ai+2. −・・−,
The label of Aa is decremented (S64). Then, the circumscribed rectangle information corresponding to the label number of Ai in the circumscribed rectangle position memory 25 is deleted, and the circumscribed rectangle information after Ai is moved one by one to the previous label (S65). S62
．． S63. S64 is preprocessing for S65, and S65
The maximum label number finally obtained by executing will directly indicate the number of circumscribed rectangles. After that, when j becomes a (S66), the process ends (
S67).

If i has not yet become a in S66, the value of i is incremented (568) and the process returns to S62.

In the process 341 in FIG. 4, A1 in the label storage register 24 is input, and the circumscribed rectangle information of the label number A1 in the circumscribed rectangle position memory 25 is compared with the current position (x, y).
Minimum x, y coordinates and maximum X.

Save the information in the memory location labeled A1 with the y coordinate as the start point and end point coordinate respectively (if the value of a is 1 at 83)
In the case of , Aa = A1, and when the value of a is 2 or more, 84
0 for AI, A2. . . . are rearranged in ascending order of Aa and label number.

In the process of 343 in FIG. 4, the value of a becomes 0 in S39, and for those subjected to the process of S42, the newly added label number is determined to have a value of 1 or 2 or more in S39. be done. For those processed in S41, the value of A1 is stored at the current position (x+n1) of the label storage line memory 23-1.

As explained above, this embodiment has the following advantages.

The mask setting means 21 scans the binary image obtained by binarizing and sets a mask. Thereafter, the circumscribed rectangle detection means 22 determines that all black pixels appearing within the mask are in the same circumscribed rectangular area according to the size and shape of the set mask, and judges connectivity each time while scanning the binary image. death,
Label storage line memory 23, label storage register 24
Then, circumscribed rectangles are sequentially formed and stored using the circumscribed rectangle position memory 25, and the circumscribed rectangles are extracted. Therefore, processing can be performed pixel by pixel (line by line), and the capacity of the memory 23, 25 and register 24 used for processing can be reduced, and the processing time can be shortened. Furthermore, by arbitrarily setting a mask in one scan, it is possible to accurately extract circumscribed rectangles under various conditions.

Note that the present invention is not limited to the illustrated embodiment, and the flowchart showing the operations of the mask setting means 21, the circumscribed rectangle detection means 22, etc. can be modified in various ways.

(Effects of the Invention) As described above in detail, according to the first invention, a binary image obtained by binarization is input, a mask is arbitrarily set by the mask setting means, and circumscribed 7 rectangle detection is performed. By using this method, connectivity is determined each time a binary image is scanned, and circumscribed rectangles are sequentially formed and saved, making it possible to process, for example, pixel by pixel (line by line), and the memory used for processing. Not only can the capacity be reduced, but also the processing time can be shortened.

Second. In the third invention, for example, by appropriately setting the size and shape of the mask in one scan, circumscribed rectangles under various conditions can be easily extracted.

In the fourth to eighth inventions, by using the label storage line memory, the label storage register, and the circumscribed rectangle position memory, there is an effect that the extraction process of the circumscribed rectangle can be further simplified and simplified.

[Brief explanation of the drawing]

FIG. 1 is a functional block diagram of an embodiment of the present invention, and FIG. 2 (a
) to (e) are diagrams showing examples of mask settings in the case of scanning direction 1 in FIG. 1, and FIG. 3(a). (b) is a diagram showing an example of mask setting in the case of scanning direction 2 in FIG. 1, FIG. 4 is a flowchart for detecting a circumscribed rectangle in FIGS. 6 is a flowchart of step 840 in FIG. 4, and FIG. 7 is a flowchart 1 to S40 in FIG. 4 for explaining detection.

Claims (1)

[Claims] A binary image obtained by 1 and binarization is input to a mask setting means, and the mask setting means determines that all black pixels appearing in a certain area are in the same circumscribed rectangular area. After arbitrarily setting a mask indicating a certain area to be scanned, the circumscribed rectangle detection means scans the binary image according to the mask set by the mask setting means, judges the connectivity each time, and sequentially detects the circumscribed rectangle. A method for cutting out a circumscribed rectangle of a document image, the method comprising forming and storing a circumscribed rectangle. 2. The method for cutting out a circumscribed rectangle of a document image according to claim 1, wherein the mask setting means determines the size of the mask if a portion of unscanned pixels beyond the position of interest is not included within the range of the mask. - A method for cutting out a circumscribed rectangle from a document image configured to arbitrarily set the shape. 3. The method for cutting out a circumscribed rectangle of a document image according to claim 1, wherein the circumscribed rectangle detecting means determines that all black pixels appearing within the mask range created by the mask settings are the same circumscribed rectangular area; A method for cutting out a circumscribed rectangle of a document image, in which the circumscribed rectangle is sequentially formed and saved while scanning a binary image. 4. In the method for cutting out a circumscribed rectangle of a document image according to claim 1, when the circumscribed rectangle detection means is executed, a label storing circumscribed rectangle information corresponding to the size of the mask is stored in a label storage line memory. , a document image in which the circumscribed rectangles are sequentially formed and saved by storing different labels existing within the arbitrarily set mask in a label storage register and further storing the circumscribed rectangle information in a circumscribed rectangle position memory. How to cut out the circumscribed rectangle. 5. The method for cutting out a circumscribed rectangle of a document image according to claim 4, wherein the label storage line memory stores n1 of the set mask.
, n2 and the size of the image are set as one label storage line memory, and the position of the label storage line memory corresponding to the coordinate values (x, y) of the binary image is determined by ( x+n1) A method for cutting out a circumscribed rectangle of a document image. 6. The method for cutting out a circumscribed rectangle of a document image according to claim 4, wherein the label storage line memory and the label storage register store a label storing circumscribed rectangle information to which the pixel belongs in the case of a black pixel; In this case, how to cut out the circumscribed rectangle of a document image by saving the initial value as is. 7. In the method for cutting out a circumscribed rectangle of a document image according to claim 4, when the number of different label numbers stored in the label storage register is two or more, the circumscribing rectangle detecting means detects the number of different label numbers stored in the label storage register. The first step is to integrate and save circumscribed rectangle information corresponding to a number of labels (where a is an arbitrary number), and if the number is 1 or 2 or more, the coordinates of the current position of interest are The second step is to integrate the value (x, y) and the circumscribed rectangle information corresponding to its label number and save it at the same label position, and if the number is 0, the coordinate value (x, y) of the target position is newly integrated. y) A third step of adding information as circumscribed rectangle information. 8. In the method for cutting out a circumscribed rectangle of a document image according to claim 7, the a number of labels stored in the label storage register are arranged in ascending order of A1, A2, . . . Aa, and the label information of A1 and other Find new information by comparing with the label information, save the result to the label of A1, delete the information of the integrated label, move the information of subsequent labels to the previous label, and A method for cutting out a circumscribed rectangle of a document image, in which the contents of the label storage line memory and the label storage register are rewritten correspondingly, so that the maximum label always indicates the number of circumscribed rectangles as it is.