JPH07120390B2 - Character cutout method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a character cutting method, and more specifically, it reads characters written on a form, and a character pattern string based on the read characters is read for each character area. The present invention relates to a character cutting method for separating and extracting.

(Prior Art) In an optical character recognition device (hereinafter abbreviated as OCR), characters written on a form are scanned line by line, an optical signal is converted into an image signal by a photoelectric converter, and the image signal is stored in a line buffer. The line buffer is read sequentially and the character pattern string is separated for each character area, and recognition is performed based on the separated character pattern. Therefore, the character segmentation method that extracts one character area from the character pattern string is the OCR performance. Greatly affect the.

In the OCR, one line is scanned from the upper end to the lower end of the line buffer in which the character string is stored, and the character pattern of the line buffer is read out by sequentially moving the line in a direction perpendicular to this scanning. In addition, a histogram is created by counting black dots (black dots in a character portion and white dots in a background portion) during scanning of one row, and the area of one character is determined by referring to the black dot histogram.

However, in the case of handwritten characters, the characters are adjacent to each other because the characters are being written with a slant, or because the characters have entered outside the character entry frame, or because the writer has partially spelled the characters. Characters overlap and 2
There is a problem that a character pattern of more than characters is cut out as one character.

To solve this problem, Japanese Patent Application No. 60-36573 (Japanese Patent Laid-Open No. 61-196381) previously proposed by the present applicant creates a black dot histogram for a character pattern string by a histogram creating means, The width from the start point to the end point of the black dot histogram is detected by the detecting means by a predetermined threshold value. Determine how many characters correspond to the width of this black dot histogram, and
If the number of characters is greater than or equal to the number of characters, the character circumscribing frame of the character pattern string is detected, and the character circumscribing frame is divided into areas of a predetermined number of divisions in the horizontal direction. In the above, the starting point and the ending point are detected again by the histogram creating means and the detecting means, and the cut-out position of each character pattern is determined for each area of the division number.

FIG. 6 shows an example of the pattern of the conventional character cutting method described above. 100 is the outer frame of the patterns 101 and 102,
103 is a horizontal bisector of the circumscribing frame 100 (hereinafter referred to as the C axis),
Reference numerals 104 and 105 denote blocks of the black dot histogram of the pattern in the upper area projected on the C axis. Reference numeral 108 denotes the bottom side of the circumscribing frame 100 (hereinafter referred to as the E axis). 106 and 107 are E axis 1
It shows the block of the black dot histogram of the pattern between the C and E axes projected on 08. Blocks 109 and 110 are 1
The midpoints of 04,105 and 106,107 are shown. The circumscribing frame 100 is divided into two regions by the horizontal bisector line (C axis) 103, and a black dot histogram in the vertical direction is created in each region.
4 to 105 and 106 to 107 are detected, and the midpoints 109 and 110 between the blocks in each area are set as cutout points, and as shown in FIG.
-A 'was set as the cutout position.

(Problems to be Solved by the Invention) However, in the above-described conventional character segmentation method, the block of the black dot histogram is 1 in the horizontally divided area of the circumscribing frame.
If there is more than one area that can be detected only one, it is impossible to cut out one character correctly.

Here, when the letters are more complicated and overlapped as an example, some of the letters are greatly intruding into the adjacent letters because the letters may be larger than the entry frame or may be bounced off due to the habits of the person who fills out the form, habits, etc. Is shown in FIGS. 7 (a) and 7 (b). The figure (a) shows the case where the area in the character circumscribing frame 200 including the patterns 201 and 202 is equally divided into two by the C axis 203, and the figure (b) shows the area in the same character circumscribing frame 200 by the C ₁ axis 203 and C ₂ axis
Shown is the case of dividing into three equal parts in 204. E axis 210 in any case
The black dot histogram of the pattern projected on is block 209.
However, there is a problem that only one is detected, and accordingly, the correct cutout position cannot be detected. Furthermore, if the number of horizontal divisions of the character circumscribing frame is increased, the processing time also increases.

An object of the present invention is to solve these problems, and an object thereof is to provide a highly accurate character cutting method with a simple configuration.

(Means for Solving Problems) In order to solve the above problems, the present invention stores a quantized character pattern string obtained by photoelectrically converting a written character string on a form in a line buffer memory. Then, in the character cutting method for separating and extracting the character pattern for each character from the character pattern sequence, the line buffer memory is scanned in the column direction of the character pattern sequence for each column to create a black dot histogram in the column direction, A determination step of comparing a block width of a horizontal block in which black dot histograms larger than a first threshold value continuously continue for a second threshold value or more with a third threshold value and determining how many characters the width corresponds to; A character pattern string corresponding to the block width of the horizontal block is scanned row by row in a row direction to create a row-direction black dot histogram, and black dot histograms larger than the first threshold are consecutive. And a circumscribing frame detecting step of detecting a character circumscribing frame based on the coordinates of the starting point and the ending point of the vertical block continuing for the second threshold or more and the coordinates of the starting point and the ending point of the horizontal block, and the detected character circumscribing frame. A pattern storing step for holding a character pattern string in the inside, a black dot histogram creating step for creating a black dot histogram in the column direction by scanning one by one for each divided area obtained by horizontally dividing the character circumscribing frame, and the created black dot Using the histogram storage step of storing the histogram for each divided area and the stored black dot histogram, the horizontal blocks that are the continuous portions of the black dot histogram in the desired divided area are detected, and the horizontal center point of the adjacent horizontal block is detected. The step of determining the cutout point of the divided area based on the coordinates of the The divided area is redivided in the horizontal direction, and the pattern storage step, the black dot histogram creation step, the histogram storage step and the determination step are performed to perform a redivision processing for detecting a block of the redivided area. The subdivision processing step and the subdivision processing step are repeated until the cutout points of all the areas can be determined, and the character cutout position is determined based on the determined cutout points of each divided area and the horizontal division line. .

(Function) According to the present invention, since the character cutting method is configured as described above, the technical means functions as follows.

In the determination step, the line buffer memory is scanned for each column in the column direction of the character pattern column to create a black dot histogram in the column direction, and black dot histograms larger than the first threshold value are consecutively continued for the second threshold value or more. The block width of the block is compared with the third threshold to determine how many characters the width corresponds to. Next, in the circumscribing frame detecting step, the character pattern string corresponding to the block width of the horizontal block is scanned row by row in a row direction to create a black dot histogram in the row direction.
The character circumscribing frame is detected based on the coordinates of the starting point and the ending point of the vertical block and the coordinates of the starting point and the ending point of the horizontal block in which the black dot histograms larger than the threshold value continuously continue for the second threshold value or more. The pattern storing step holds the character pattern string in the detected character circumscribing frame. In the black dot histogram creating step, a black dot histogram in the column direction is created by scanning one column for each divided area obtained by horizontally dividing the character circumscribing frame. The histogram storage step stores the created black dot histogram for each divided area. The determining step uses the stored black spot histogram to detect horizontal blocks that are continuous portions of the black spot histogram in the desired divided area, and based on the coordinates of the horizontal center point of the adjacent horizontal block, the cut-out point of the divided area. To decide. Then, in the re-division processing step, the division area is re-divided in the horizontal direction with respect to the division area in which the cut-out point cannot be determined, and the pattern storage step, the black point histogram creation step, the histogram storage step and the determination step are performed to determine the re-division area. The subdivision processing for detecting the blocks is performed, and this subdivision processing step is repeated until the cutout points of all the regions can be determined. The character cutout position is determined based on the finally determined cutout point of the divided area and the horizontal dividing line.
Therefore, even when a part of a character is deeply inserted between adjacent characters, the character can be cut out with high accuracy.

(Embodiment) An embodiment of the present invention will be described with reference to FIGS. 1 to 5. FIG. 1 is a block diagram showing an embodiment of the present invention. In the figure, 300 is an image signal from an unillustrated photoelectric conversion unit, 301 is a line buffer, 302 is a pattern buffer, 303 is an address counter in the X direction of the pattern buffer, 304 is an address counter in the Y direction, and 305 is a line buffer 301. A switching circuit for switching between the output and the output of the pattern buffer 302, 306 a black dot histogram creating circuit of the line buffer 301 or the pattern buffer 302, 307 a one-line histogram memory for storing the black dot histogram of the line buffer 301, and 308 a black dot histogram of the pattern buffer 302. Pattern histogram memory, 309
Is a switching circuit for switching between the one-row histogram memory 307 and the pattern histogram memory 308, 310 is a black dot histogram operation circuit, 311 is a black dot histogram block and a block detection circuit for detecting a character circumscribing frame, and 312 is a control circuit. Further, in the figure, a single arrow indicates a data flow, and a double arrow indicates a control signal of the control circuit 312.

The operation of this embodiment will be described below with reference to the block diagram of FIG.

The character string on the form is converted into a binarized image signal 300 by a photoelectric converter and stored in the line buffer 301. The following processing is performed under the control of the control circuit 312. The control circuit 312 reads out the image signal stored in the line buffer 301 from the head position of the line buffer 301 in units of one column,
The sequence is further advanced, and the process ends when all the character pattern data for one line is read. Further, in the control circuit 312, the switching circuit 305 is switched to the line buffer 301, and the line buffer
At the same time as reading the pattern data in units of one column from 301, the black dot histogram creation circuit 306 is activated. The black dot histogram creation circuit 306 creates the black dot histogram of the column by counting the number of black dots being read in the column, and stores it in the one-row histogram memory 307. The above processing is repeated, and the processing is ended when the black dot histograms of all rows are stored in the one row histogram memory 307. The one-row histogram memory 307 has a capacity capable of storing a black dot histogram corresponding to the total number of columns of the line buffer 301.

After the black dot histogram for one line is created, the one line histogram memory 307 is read from the beginning and the block is detected by referring to the black dot histogram. The control circuit 312 switches the switching circuit 309 to the one-row histogram memory 307 and activates the block detection circuit 311. Block detection circuit 311 is 1
Black dot histograms are sequentially read from the line histogram memory 307, the black dot histogram is compared with a threshold value α (α: constant, but α = 1 in this embodiment), and if the black dot histogram is large, a candidate for the start point of a character block is obtained. Then, the storage addresses of the black dot histograms are sequentially updated, and the columns in which the read black dot histograms are larger than the threshold value α are counted,
When β (β: constant, but β = 2 in this embodiment) columns are continuous, the starting point candidate is set as the starting point. Further, the column is continuously updated, and after the start point is detected, the row in which the black dot histogram becomes smaller than the threshold value α is set as the end point, and the area of the horizontal block indicated by the length from the start point to the end point is detected. Next, the control circuit 312 determines the thresholds γ ₁ and γ ₂ (γ ₁ and γ ₂ are constants in the present embodiment, which are obtained from the average width of the character to be read, based on the detected horizontal block length. In γ ₁ = 75, γ ₂ = 125).
Then, when the length W of the block is smaller than the threshold γ ₁ , the block is determined to be one character, and γ ₁ ≦ W ≦ γ ₂
When it is, it is judged as two characters, and when W> γ ₂ , it is judged as 3 characters.
It is judged that it is more than the character. In addition, the control circuit 312 determines the block and then detects the block.
311 detects the circumscribing frame. The circumscribing frame is detected as follows. That is, the portion of the character string pattern corresponding to the start point to the end point of the horizontal block is scanned line by line in the row direction to create a black dot histogram in the row direction and store it in a temporary histogram memory (not shown). The same process as above is performed by sequentially reading out the black dot histograms in the row direction,
The area of continuous vertical blocks of the black dot histogram satisfying a predetermined condition is detected, and the coordinates of the circumscribing frame are obtained from the start point and end point of this block and the start point and end point coordinates of the horizontal block already obtained. Further, when the coordinates of the circumscribed frame are detected, the control circuit 312 transfers the character pattern indicated by the circumscribed frame from the line buffer 301 to the read pattern buffer 302. Here, if W> γ ₂ , that is, if the block is determined to be 3 characters or more, the cutout process is performed from the start point to γ ₂ to divide the first and second characters, and the resulting cutout point is used as the start point. The cut-out process is performed from the starting point to γ _2, and the cut-out process is further performed such that the second character and the third character are further divided and the process is performed up to W. Here, as shown in FIG. 4 which will be described later, the left end of the upper side of the character circumscribing frame is the origin, the lower side position is PB, and the right side position is PR.

Next, the processing of a character that has been determined to be two or more by the above character determination will be described with reference to FIG.

The control circuit 312 switches the switching circuit 305 to the pattern buffer 302 and activates the black dot histogram creation circuit 306. The black dot histogram creation circuit 306 creates a black dot histogram in the vertical direction within the character circumscribing frame for the character pattern string stored in the pattern buffer 302. The details will be described below with reference to FIGS. 2 and 3.

FIG. 2 is a block for creating a vertical black dot histogram of the pattern buffer 302. In the figure, 308 is the same pattern histogram memory as in FIG. 1, 400 is the black dot histogram of the black dot histogram creating circuit 306, and 401 is the output of the X address counter 303 of the pattern buffer 302. 402 is an area address counter that horizontally divides the character circumscribing frame and indicates an address from the upper side of the character circumscribing frame of the divided area, 403 is the output of the Y address counter 304 of the pattern buffer 302, and 404 is provided from the detection circuit 312. Division information of the character circumscribing frame, 405 is a division position designating circuit that calculates the division position based on the division information 404, and 406 is a Y address counter
A division position coincidence detection circuit that detects a coincidence between the output 403 of 04 and the output of the division position designation circuit 405, 410 is an output of the pattern histogram memory 308, and 411 is a coincidence signal that notifies the control circuit 312 that the division positions coincide. ing. Reference numeral 412 is a control signal (address) from the control circuit 312 that controls the area address counter 402.

Next, FIG. 3 shows the structure of the pattern histogram memory 308 shown in FIG. 1 and FIG. 501 is an area for character circumscribing frame detection, 502, 503 are temporary areas for dividing areas (1) and (2) which are areas for calculating a black point histogram and a block, and 504 to 509 are black point histograms for predetermined divided areas. It shows the area to be used.

A method of creating a black dot histogram in the vertical direction for a character pattern string in a character circumscribing frame will be described with reference to FIGS. 1 and 2. When the character circumscribing frame is detected, the control circuit 312 calculates a predetermined division position in the Y address direction. In this embodiment, the number of divisions is 6. Next, the control circuit 312 sets a predetermined address (412) in the area address counter 402 shown in FIG. 2, sets the division position information 404 in the division position designation circuit 405, and then sets the black dot histogram creation circuit 306.
To start. The black dot histogram creation circuit 306 initializes the X address counter 303 and the Y address counter 304, increments the Y address counter 304 to count the black dots of the pattern, and when the Y address reaches the lower side of the character circumscribing frame. The count result of the black dots is stored in the pattern histogram memory 308, the X address counter 303 is incremented, and the Y address counter 304 is cleared. The above processing is repeated until the X address counter 303 reaches the right side of the character circumscribing frame. The coordinates of the pattern buffer 302 are in the fourth quadrant.

Next, a method for creating a black dot histogram for each divided area of the character circumscribing frame will be described with reference to FIGS. The Y address counter 304 is incremented by the black dot histogram creation circuit 306 and the division position given by the division position designation circuit 405 and the Y address counter 304 are divided position coincidence detection circuit 406.
When a match is detected by, the number of black points counted up to that point is stored in the position indicated by the area address counter 402 and the X address counter 401 of the pattern histogram memory 308. At the same time, the area address counter 402 is incremented. The area address counter 402 is initialized at the same time when the Y address counter 304 is initialized. As described above, the black dot histogram for each divided area is created and stored in the divided area (1) 504 to the divided area (6) 509 of the pattern histogram memory 308 shown in FIG.

When the black dot histogram for each divided area is created, the control circuit 312 in FIG.
Start and create a black dot histogram of the upper and lower divided areas when the character circumscribing frame is divided at the center. The creation method is the black dot histogram itself stored in the divided area (3) 506 shown in FIG. 3 for the upper side, and the lower area is from the black point histogram of the divided area (6) 509 to the divided area (3) 506. It is the rest after subtracting the black dot histogram of. That is, the black dot histogram of a certain divided area is set to Hnm
(N is the number of the divided area of 1 to 6, m is 0 to PR and PR is the right end of the character circumscribing frame), and in the obtained black dot histogram, the upper divided area is Hum and the lower side is H _L m. , Calculated by the following formula.

Hum = H ₃ m H _L m = H ₆ m−H ₃ m When the black dot histogram calculation circuit obtains the black dot histogram of the two divided areas, the temporary area (1) 502 for the divided areas of the pattern histogram memory 308 shown in FIG. 3 is obtained. And (2) 503.

Next, the control circuit 312 switches the switching circuit 309 to the pattern histogram memory 308, and activates the block detection circuit 311 for the detected black dot histogram to detect the block.
The thresholds at this time were α = 1 and β = 2. When a block is detected, the midpoint between adjacent blocks is determined and used as the cutout point. When a plurality of cutout points are detected in one divided area, the cutout point is closest to the horizontal center position of the character circumscribing frame. When the cutout points are determined in each of the upper and lower divided areas, the cutout position is determined by those cutout points. If the cut-out point is not determined even in one of the divided areas, the black-point histogram calculation circuit 310 creates a black-point histogram of the area that has not been determined and is further divided into three, and the block detection circuit 3
In 11, the block of the black dot histogram is detected in each of the three divided areas, and the cut-out point for each divided area is determined.
When all the cut-out points of the divided area are detected, the cut-out position of the character is determined by each cut-out point and the horizontal dividing line.

The pattern cutting method will be described below with reference to the pattern example shown in FIG. 4 in which the cutting position is determined. Also, the fourth
The figure shows the pattern buffer 302 in the block diagram of FIG.
The pattern and the cut-out position stored in FIG. The coordinates are such that the horizontal axis is the X axis and the vertical axis is the Y axis, and the pattern buffer 302 is located in the fourth quadrant. XM
And YM indicate the size of the pattern buffer 302,
In this embodiment, XM = YM = 128 mesh. PR and PB indicate the circumscribing frames of the patterns stored in the pattern buffer 302, and are represented by four straight lines X = 0, X = PR, Y = 0, Y = PR. In FIG. 4, 600 and 601 are patterns, and the polygonal line indicated by 602 is the cutout position. y
Divide the upper and lower divided areas by the dividing line of ₃ and detect the block of the black dot histogram of each area to find the cutout point of each divided area.The upper side is detected as x ₁ , but the lower side detects the cutout position. I can't. Therefore, the points specified in advance for the area from y ₃ to PB on the lower side
Divide into 3 by y ₄ and y ₅ . Next, for each divided area, block detection is performed by the black dot histogram of that area to determine the cutout points, and x ₂ and x ₃ are obtained. As described above, the cutout line of the character is determined as the polygonal line 602. Therefore, the character circumscribing frame and the pattern 6 within the polygon indicated by the polygonal line 602
Transfer 00,601 to the next stage.

Next, the flow of this embodiment will be described in detail based on the flow chart shown in FIG. In FIG. 5, the process starts at S700. After the read position of the line buffer 301 is set to the first column in S701 and initialization is performed, the pattern data stored in the line buffer 301 is processed in S702 via the one-column read switching circuit 305 and the black dot histogram creation circuit 306 creates a black dot histogram. It is created and stored in the one-line histogram memory 307. In S703, the completion of creation of the black dot histograms for all the lines is detected, and the process of S702 is repeated until all the lines are created. In S704, the processed characters are managed and 1
The following process is repeated until cutting out all the characters in the line is completed. In S705, the black dot histogram is read from the one-line histogram memory 307, and the start point and end point of the black dot histogram are detected and set as a block. Also, the length of the block is compared with the thresholds γ ₁ and γ ₂ to hold how many characters the block is made up of. In S706, the block detection circuit 311 detects the circumscribing frame of the block, and transfers the pattern data in the circumscribing frame to the pattern buffer 302. In S707, the pattern data of the pattern buffer 302 is transferred to the output stage if there is one character according to the judgment result of the length of the held block and the process proceeds to the next character.
If it is more than the character, the following processing is performed. In S 708, a black dot histogram in the vertical direction is created in the character pattern string in the character circumscribing frame for each specified divided area and stored in the pattern histogram memory 308. In S709, the division position of the character circumscribing frame is selected at the center, and in S710, the black dot histogram calculation circuit 310
The black dot histograms of the upper and lower divided areas divided into two are created respectively and stored in the pattern histogram memory 308. In S711, the block of the black dot histogram of the two areas created in S710 is detected, and the cutout point is determined. S712
Then, the position is determined from the cutout point of each divided area, and if it is determined in all the divided areas in S713, the cutout position of the character is detected in S714, the pattern is transferred to the next stage in S715, and the next character Go to processing. In S713, if there is a region where the cutout point has not been determined, the flow proceeds to S717, the division point for further dividing into three regions is determined, and the flow proceeds to S710 to again create a black dot histogram for each divided region, and detect the block. To determine the cut-out point of the divided area. By the above processing, the character cutout position is determined.

In the present embodiment, in the divided area where the cutout point cannot be detected, the cutout point can be detected by one reproduction division, but the cutout point detection processing is repeatedly performed for the area where the cutout point cannot be detected to detect the cutout point. By doing so, it is possible to cut out more complicated overlapping characters.

As described above, according to the present embodiment, even when the character patterns before and after overlap with the character pattern, the character pattern is not missing or part of the character pattern before and after is mixed without changing the character pattern. It can be cut out.

Further, in the present embodiment, the case where two characters overlap each other has been shown. However, even when three characters or more overlap each other, the cutout points are sequentially determined based on two characters from the beginning of the overlapped characters. Similar effects can be obtained.

(Effect of the invention) As described above, according to the present invention, a character circumscribing frame is horizontally divided to create a black dot histogram, and the cutout position is determined for each divided area. It is possible to perform highly accurate character cutout without omission or mixing of some other characters. The cut point of the divided area is not determined.Since the subdivision is repeatedly performed until the cut point is detected, even if the adjacent characters overlap in a complicated manner, the processing time is fast and the accurate cut is performed. Can be done. Since the black spot histogram for each of the divided areas is created by performing the pattern scan once for the number of areas designated in advance, it is possible to realize the character cutting with a short processing time. Further, since a highly accurate character cutout is realized by using a black dot histogram created by a process of simply counting black dots, it can be realized with a simple circuit configuration.

Further, by using the present invention, it is possible to cut out when adjacent characters overlap each other, so that the interval between the character entry frames can be reduced and the number of readable characters per line can be increased. Therefore, there is an effect that it can be applied to various types of recording media such as forms, the degree of freedom in designing forms is large, and thus an optical character recognition device with good performance can be realized.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a block diagram for creating a black dot histogram in the character circumscribing frame of the embodiment of FIG. 1, and FIG. 3 is a pattern histogram memory configuration. FIGS. 4A and 4B are diagrams showing pattern examples in which the clipping position is determined, FIG. 5 is a flow chart showing the character clipping process of the embodiment of FIG. 1, and FIG. 6 is a pattern example by the conventional character clipping method. FIGS. 7A and 7B are diagrams showing an example of a pattern that causes a problem in the conventional method. 300 …… image signal, 301 …… line buffer, 302 …… pattern buffer, 303 …… X address counter, 304 …… Y
Address counter, 305, 309 ... Switching circuit, 306 ... Black dot histogram creation circuit, 307 ... One-line histogram memory, 308 ... Pattern histogram memory, 310 ... Black dot histogram calculation circuit, 311 ... Block detection circuit, 312 ...
... Control circuit, 400 ... Black dot histogram, 401 ... X address counter output, 402 ... Area address counter, 4
03 …… Y address counter output, 404 …… Division information, 4
05: Divided position designation circuit, 410: Pattern histogram memory output, 411: Match signal.

Claims (1)

[Claims] 1. A quantized character pattern string obtained by photoelectrically converting a written character string on a form is stored in a line buffer memory, and character patterns for each character are separated and extracted from the character pattern string. In the character cutting method, the line buffer memory is set to 1 in the direction of the character pattern string.
A black dot histogram in the column direction is created by scanning for each column, and a block width of a horizontal block in which black dot histograms larger than a first threshold value continuously continue for a second threshold value or more is compared with a third threshold value. And a character step corresponding to the block width of the horizontal block in the row direction is scanned line by line to create a black dot histogram in the line direction. A circumscribing frame detection for detecting a character circumscribing frame based on coordinates of start and end points of a vertical block in which black dot histograms larger than a threshold of 1 continuously continue for the second threshold or more, and start and end coordinates of the horizontal block A pattern storing step for holding a character pattern sequence in the detected character circumscribing frame, and 1 for each divided area obtained by horizontally dividing the character circumscribing frame.
A black dot histogram creation step for creating a black dot histogram in the column direction scanned for each column, a histogram storage step for storing the created black dot histogram for each divided area, and a black dot histogram for a desired divided area using the stored black dot histogram A determination step of detecting horizontal blocks that are continuous parts and determining the cutout points of the divided areas based on the coordinates of the horizontal center points of adjacent horizontal blocks, and the division of the divided areas where the cutout points cannot be determined. A subdivision processing step of subdividing the area in the horizontal direction, performing the subdivision processing of performing the pattern storage step, the black dot histogram creation step, the histogram storage step, and the determination step to detect a block of the subdivision area, The subdivision is performed until the cut points of all areas can be determined. A character slicing method, characterized in that the processing step is repeated to determine a character slicing position based on the determined slicing point of each divided region and a horizontal dividing line.