JPH0765122A - Optical character reader - Google Patents

Abstract

PURPOSE:To improve the processing time by efficiently using one or plural recognition parts by performing character segmentation for recognizing a reread character after characters are segmented in recognized character group units. CONSTITUTION:A control part 2 segments characters 1-6 in a 1st line in order and transfers them to recognition parts #1-#3. Here, the segmentation time of one character is set to tK and the recognition time of one character is set to tR are so set that 3tK<=tR. When the character 6 is transferred to the recognition part #3 and begins to be recognized, the control part 2 starts segmenting characters in a 2nd line. Once the recognition of the six characters 1-6 in the 1st line ends, the control part 2 decides whether a read is made by 1st scanning, namely, whether the characters can correctly be read. When there is a character which can not be read, 2nd scanning is performed for the character which can not be read and the reread character is segmented for recognition.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reader for optically reading image data and converting it into an electric signal.

[0002]

2. Description of the Related Art FIG. 7 is a block diagram showing a structural example of a conventional optical character reading device. 7, in this character reading device, a control unit 102 and a multi-valued image memory 10 connected to each other via an address data bus 101.
3, recognition unit # 101, recognition unit # 201, recognition unit # 301
The multi-valued image memory 103 is connected to an optical system 104 which optically reads an image of the form 105 and converts it into an electric signal.

In the character reading apparatus thus constructed, when the image of the form 105 is optically read by the optical system 104 and converted into an electric signal, this is stored in the multi-valued image memory 103. Further, the signals stored in the multi-valued image memory 103 are transmitted to the control unit 102.
The characters are cut out for each character, and the cut out characters are transferred to the plurality of recognition units # 101 to # 301 for each character for recognition.

FIG. 8 shows an example of the form shown in FIG. 7, and FIG. 9 shows a time chart until the character cut out from the multi-valued image memory 103 is recognized in the conventional character reading apparatus. . Therefore, in the character reading apparatus shown in FIG. 7, a processing procedure until the character cut out from the multi-valued image memory 103 is recognized will be described below with reference to FIGS. 8 and 9.

First, the control unit 102 cuts out the first character "0" in the first line 101L by the character cutting process and transfers it to the recognition unit # 101. Then, the recognition unit # 10
Recognition at 1 is started. At the same time, the first line 1
The second character “1” of 01L is cut out, and this is the recognition unit #
Then, the recognition unit # 201 starts recognition of the character. Next, the third character "2" in the first line 101L is cut out, transferred to the recognition unit # 301, and then recognized.
At 301, character recognition is started. Furthermore, the recognition unit # 3
The recognition in 01 is started, and the first in the recognition unit # 101 is started.
When the character recognition is completed, the fourth character "3" in the first line 101L is cut out, transferred to the recognition unit # 101, and the recognition unit # 101 starts the character recognition. Less than,
Similarly, the fifth character "4" and the sixth character "5" are recognized part #.
201 and # 301 respectively. This process will be referred to as "first scan" in the following description.

When the recognition of the last character of the first line 101L (the sixth character "5" in this example) is completed, the first line 10L
Perform 1L editing processing. Here, for example, when the character “0” and the character “4” are not correctly read (hereinafter referred to as “non-read”), the control unit 102 changes the slice in the density direction from the multi-valued image memory 103. Letter "0"
Is cut out, and the cut out character “0” is transferred to the recognition unit # 101. Then, the recognition unit # 101 starts recognition. At the same time, the character "4" is cut out, and the recognition unit # 201 starts recognition. This process is referred to as "second scan" in the following description.

When the recognition of the characters "0" and "4" in the second scan is completed, the editing process is started. Here, if the characters "0" and "4" have been correctly read (hereinafter referred to as "correct reading"), the first scan of the second line 102L is started. If the first scan on the second line is correct reading, the first scan on the third line 103L is performed.
The scan is started and similar processing is performed. Thereafter, similar processing is performed for each row.

That is, in this conventional optical character reader, the correctness / non-reading is judged after the completion of the first scanning of each line, and if the reading is not performed, the second scanning of the corresponding line is carried out. After the second scan is finished, the process of determining whether the line is correct or unreadable is repeated until a predetermined scan (usually about the fifth scan), and then the first scan of the next line is performed. The recognition rate is improved by taking such a procedure. If the character cutout time is t _k and the character recognition time is t _R , the character cutout time t _k is shorter than the character recognition time t _R , that is, (t _k ≤t _R ). Multiple recognizers #
Throughput is improved when processing is performed using 101 to # 301.

[0009]

As described above, in the conventional optical character reading device, in order to improve the recognition rate, the correctness / non-reading is determined after the completion of the first scan of each line,
If there is a non-read, the second scan of the corresponding line is performed, and after the end of the second scan, the scan process is repeated a plurality of times so as to judge again whether the line is correct or not. After that, the first scan of the next line is performed. However, in this method, as in the case of the recognition unit # 103 at the time of the second scan of the first line, other recognition units # 101,
While # 102 repeats the scanning process, there is also a recognition unit that is not used for the recognition process and rests, and there is a problem that a plurality of recognition units # 101 to # 103 are not used effectively, However, there is a problem in that a waiting time for cutting out occurs and the processing time is reduced.

The present invention has been made in view of the above problems, and an object thereof is to read an optical character in which one or a plurality of recognition units can be efficiently used to improve the processing time. To provide a device.

[0011]

According to the present invention, there is provided a photoelectric conversion unit for optically reading a form and converting it into a multivalued image, and an image of the form photoelectrically converted into a multivalued image. And an image memory for reading out a form image while switching a plurality of slices in the density direction from the image memory, and a character cutout portion for cutting out a character for each character, and the character cut out by one or a plurality of recognition portions. In an optical character reading device, which comprises a recognition unit for recognizing, performs character cutting in the slice, and recognizes a recognized character group unit such as a line unit or a field unit for rerecognizing by switching a predetermined slice based on the recognition result. After the character segmentation of the recognized character group unit is completed, the subsequent character segmentation of the recognized character group unit is executed, and After the recognition of the unit is completed, the character in the recognized character group unit is judged to be correct or unreadable, and when a character that needs to be reread occurs, the character is cut out for recognition of the reread character. This is achieved by providing a control means that is performed after cutting.

Further, according to the present invention, the object is to store a plurality of sheets of photoelectric conversion units for optically reading a form and converting it into a multi-valued image, and a plurality of images of the form photoelectrically converted into a multi-valued image. An image memory, and a character cutout unit that reads a form image while switching a plurality of slices from the image memory and cuts out a character for each character.
A recognition unit that recognizes the cut-out characters by one or a plurality of recognition units, cuts out characters in predetermined slices, and switches the predetermined slices based on the recognition result to re-recognize line units or field units. Is an optical character reading device for recognizing the recognized character group unit, and immediately after the character is cut out in the recognized character group unit, the recognized character of the other form stored in the next same form or in the image memory. In addition to executing character segmentation for each group,
After the recognition of each recognition character group unit, it is determined whether the recognition character group unit is correct or non-readable, and if a reread character occurs, a character cutout for recognition of the reread character is made in the same form or in the image memory. This is achieved by providing a control means for performing after the character is cut out in units of a recognized character group with another form stored in.

[0013]

According to this configuration, the character cut-out of the next line is executed immediately after the character cut-out of each line, and after the character recognition of the line is completed, it is determined whether the line is correct or not, and when re-reading occurs, After cutting out a line for recognizing reread characters, or after cutting out a line in the same form or in a unit of a recognized character group in another form stored in the image memory, the line is cut out. Since this is performed, the character slicing unit and the recognition unit operate efficiently, and the processing time can be significantly improved.

[0014]

Embodiments of the present invention will be described in detail below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of an optical character reading device according to the present invention. In FIG.
In this character reading device, a control unit 2, a multi-valued image memory 3, which are connected to each other via an address data bus 1,
The multi-valued image memory 3 is composed of a comparator 4, a pattern memory 5, a recognition unit # 1, a recognition unit # 2, a recognition unit # 3, etc., and optically reads the image of the form 6 into an electrical signal. The optical system 7 is connected.

In the character reading device thus constructed, when the image of the form 6 is optically read by the optical system 7 and converted into an electric signal, this is stored in the multi-valued image memory 3. .

Further, the signal stored in the multi-valued image memory 3 is compared with a predetermined value designated by the control unit 2 by the comparator 4 to be a binary image, which is then transferred to the pattern memory 5 of the control unit 2 by 1.5. This control unit 2
Then, the characters are cut out character by character, and the cut out character patterns are transferred to the plurality of recognition units # 1 to # 3 character by character to start recognition.

At the same time, the control unit 2 transfers the next character from the multi-valued image memory 3 to the pattern memory 5, cuts out the next character, and the cut-out character from the control unit 2 is recognized by the recognition unit # 2. The recognition unit # 2 starts recognition. Similarly, the character cut out next is transferred to the recognition unit # 3, and the character is recognized.

FIG. 2 shows an example of the form shown in FIG. 1, and FIG. 3 shows how the characters read from the multi-valued image memory 3 are recognized in the character reading apparatus of this embodiment shown in FIG. It shows a time chart. Therefore, in the character reading apparatus shown in FIG. 1, a processing procedure until the character pattern cut out from the multi-valued image memory 3 is recognized will be described below with reference to FIGS.

First, in the explanation, the reference numerals in FIG. 3 will be explained.
2 is the number of recognitions of the recognition unit # 1, S13 is the number of recognitions of the recognition unit # 2, S14 is the number of recognitions of the recognition unit # 3, and S15 is J of the I-th row.
The number of edit processing based on the scan result, that is, the recognition result of (I, J), S16 is the number of lines being processed, and S17 is the number of recognition scans.

First, the control unit 2 cuts out the first character "1" in the first line 1L of the form 6 by the character cutting process, and transfers it to the recognition unit # 1. Then, recognition by the recognition unit # 1 is started. At the same time, the second character “2” on the first line 1L is cut out, and this is the recognition unit # 2.
The recognition unit # 2 starts character recognition. Next, the third character "3" on the first line 1L is cut out, transferred to the recognition unit # 3, and the recognition unit # 3 starts recognition of the character. Further, when the recognition unit # 3 starts recognition and the recognition unit # 1 finishes recognition of the first character, the fourth character “4” on the first line 1L is cut out, and this is recognized by the recognition unit # 1.
Then, the recognition unit # 1 starts recognition of characters.
Similarly, the fifth character "5" and the sixth character "6" are 1
Each character is cut out and recognized by the recognition units # 2 and # 3.

Here, the extraction time for one character is t _k , 1
When the character recognition time is t _R , (3t _k ≤t _R ) is set, and the recognition unit performs the recognition process most efficiently when the recognition unit has three configurations.

The cutout of the final character "6" of the first scan of the first line 1L is completed, and the character "6" is sent to the recognition unit # 3.
When the character is transferred and the recognition of the character “6” is started, the character cutout unit starts to cut out the character on the second line 2L, and the recognition units # 1 and # 2 start recognition on the second line 2L. Then, the recognition unit # 3 ends recognition of the final character “6” on the first line 1L immediately before the end of the cutout of the character “9” on the second line 2L, and continues two lines after the end of the character “6”. The recognition of the character "9" of the eye 2L will be started.

Further, at the same time, the recognition of the six characters "1" to "6" in the first line L1 is completed, so the control unit 2
Then, the editing process (S15) is started. Here, the process of editing (1, 1) is started, and it is determined whether or not the data was correctly / unread in the first scan, that is, whether or not it was correctly read. Then, the characters which are not correctly read and are determined to be unreadable (in this embodiment, the characters "2", "4", "5", "6")
4), the second scan is performed on the character that is determined to be unreadable.

Here, at the end of cutting out the last character "2" of the second line 2L, the recognition result of the first scan of the first line L1 is completed, and the second scan character of the first line 1L is determined. Therefore, the cutout of the second scan character on the first line 1L is started immediately after the cutout of the final character "2" on the second line 2L.

In the second scan of the first row 1L, the comparator 4 compares the multi-valued image of the first row 1L with a predetermined value different from the first scan designated by the control unit 2 in the figure. The binary image is transferred to the pattern memory 5 in the same manner as in the first scan, the control unit 2 cuts out each character, and the second scan of the first line 1L is performed.

On the other hand, the recognition of the final character "2" of the first scan in the second line 2L is completed by the recognition unit # 3 immediately before the cutting out of the character "5" of the second scan in the first line 1L. Next, the edit process (S15) is started, and the edit (2, 1) is executed here to determine the correctness / non-readability in the first scan of the second line 2L. If there is a character that is determined to be unreadable (two characters "7" and "8" in this embodiment), the second scan is performed on the character that is determined to be unreadable.

If the selection of the second scan character of the second line 2L is completed before the cutout of the final character "6" of the second scan of the first line 1L is completed (this embodiment is completed. ) After cutting out the last character of the second scan on the first line 1L, the character cutting of the second scan on the second line 2L is started.

On the other hand, if the selection of the second scan character of the second line 2L is not completed before the cutting out of the final character "6" of the second scan of the first line 1L, the third line is selected. Start cutting 3L.

Then, while cutting out the first scan of the third line 3L, the editing (1, 2) of the second scan of the first line 1L and the editing (2, 2) of the second scan of the second line 2L are started. Then, it is determined whether the reading is correct or not, and if there is reading, the slice in the density direction is changed and the third and fourth scans are performed. Then, after performing a predetermined number of scans (for example, four scans), the unread character is regarded as unreadable and the recognition of the line is ended. After that, the same processing is repeated until the recognition of all the characters in the form 6 is completed.

In the case of this embodiment, the first line 1
Correct reading / non-reading judgment in editing (2, 2) of the second scan of L and editing (2, 2) of the second scan of 2L of 2L is the case of correct reading. The recognition of the second line 2L ends with the second scan.

Therefore, according to the optical character reader of the above embodiment, the character cut-out of the next line is executed immediately after the character cut-out of each line, and after the character recognition of that line is completed, it is judged whether the line is correct or not. , When re-reading occurs, the line is cut out after the line is cut out, so the character cutout unit and the recognition unit operate efficiently, and it is possible to greatly improve the processing time. .

In the above embodiment, the reading unit for determining that repeated scanning is required is a line unit, but this may be a unit of a recognized character group such as a field unit or a unit of a plurality of lines. It is a thing.

It is also possible to determine whether or not a repeated scan is performed for each row, and then perform the second scan of all rows after performing the first scan of all rows.

Further, the number of recognition units may be one or more, and if the character extraction time t _k and the recognition time t _R are (nt _R ≤t _k ), n recognition units are provided. And the most efficient.

Further, when the repetitive scanning is not performed, the number of scans corresponds to 1. In this case, a multivalued image is not necessary and a binary image may be used.

FIG. 4 is a block diagram showing another embodiment of the optical character reader according to the present invention. In FIG.
Those denoted by the same reference numerals as in FIG.

4, in this character reading device, the control unit 2, the multi-valued image memory 23 with 8 gradations, the comparator 4, the pattern memory 5, and the recognition unit # which are connected to each other via the address data bus 1. 1, recognition unit # 2,
The multi-valued image memory 23 is connected to an optical system 7 for optically reading the image of the form 6 and converting it into an electric signal. further,
The multi-valued image memory 23 has two banks, a bank a and a bank b.

In the character reading device thus constructed,
The form 6a or the form 6b is conveyed to the optical system 7 by a conveying means (not shown), and the form 6a or the form 6b is conveyed.
The image of the form 6a in this embodiment is optically read by the optical system 7 and converted into an electrical signal. Then, this image signal is stored in the multi-valued image memory 23 having a 2-bank structure (bank a, bank b).

Then, the image signal stored in the multi-valued image memory 23 is compared with a predetermined value designated by the control section 2 by the comparator 4 to be a binary image,
About 1.5 characters are transferred to the pattern memory 5 of the control unit 2,
The control unit 2 cuts out each character, and the cut-out character pattern is transferred to the plurality of recognition units # 1 to # 3 for each character to start recognition.

At the same time, the control unit 2 transfers the next character from the multi-valued image memory 23 to the pattern memory 5, cuts out the next character, and the cut-out character is recognized by the control unit 2 from the recognition unit. It is transferred to # 2, and the recognition unit # 2 starts recognition. Similarly, the character cut out next is transferred to the recognition unit # 3, and the character is recognized.

FIG. 5 shows the two forms 6a and 6b shown in FIG.
FIG. 6 shows a time chart until the character cut out from the multi-valued image memory 23 is recognized in the character reading apparatus of the present embodiment shown in FIG. Therefore, in the character reading device shown in FIG.
A processing procedure until the character pattern cut out from the multi-valued image memory 23 is recognized will be described below with reference to FIGS.

First, in the description, the reference numerals in FIG. 3 will be explained.
2 is the number of recognitions of the recognition unit # 1, S23 is the number of recognitions of the recognition unit # 2, S24 is the number of recognitions of the recognition unit # 3, S25 is the Kth scan of the Jth row of I, that is, (I, J, K). The number of editing processes according to the recognition result of S, the number of lines being processed, S27, the number of scans for recognition, and S28 are distinctions between the forms 6a and 6b (a is the form 6a and b is the form 6b).

First, the control unit 2 cuts out the first character "0" of the form 6a stored in the bank 0 of the multi-valued image memory 23 in the first line 1L of the form 6a by the character cutting process. , And transfers this to the recognition unit # 1. Then, recognition by the recognition unit # 1 is started. Also,
At the same time, the second character "1" in the first line 1L is cut out, transferred to the recognition unit # 2, and the recognition unit # 2 starts recognition of the character. Next, the third character “2” on the first line 1L is cut out, transferred to the recognition unit # 3, and the recognition unit # 3 starts recognition of the character. Here, when the extraction time for one character is t _k and the recognition time for one character is t _R , this corresponds to (3t _k ≤t _R ), and the recognition unit most efficiently performs the recognition process when the recognition unit has three configurations.

The cutout of the final character "3" of the first scan in the first line 1L is completed, and the character "3" is sent to the recognition unit # 1.
When the character is transferred and the recognition of the character “3” is started, the character cutout unit starts to cut out the character on the second line 2L, and the recognition units # 2 and # 3 start recognition on the second line 2L. Then, the recognition unit # 1 finishes recognizing the last character “3” on the first line 1L immediately before the cutout of the character “6” on the second line 2L, and continues to 2 after the end of the last character “3”. The recognition of the character "6" on the second line 2L will be started.

At the same time, since the recognition of the four characters "0" to "3" in the first line L1 is completed, the control unit 2 starts the editing process (S25). Here, edit (1, 1,
Step 1) is undertaken, and it is determined whether the first scan is correct or unreadable, that is, whether or not it was correctly read. Then, the character which is determined to be unreadable (in this embodiment, the characters "0" and "2" are 2
2), the second scan is performed on the character that is determined to be unreadable.

Next, when the cutting out of the final character "7" of the second line 2L is completed, the recognition result of the first scan of the first line L1 of the form 6a is completed, and the second scan character of the first line 1L is Since it has been decided, the cutout of the control unit 2 starts the cutout of the first row 1L of the second form 6b stored in the bank b of the multi-valued image memory 23.

The first scan of the first row 1L of the form 6b is performed in the same procedure as the cutting out of the form 6a at the first scan. Then, the four characters (“8”, “9”, “0”, “1”) of the first line 1L of the form 6b are cut out by the control unit 2 and recognized by the recognition units # 1 to # 3. It

The recognition by the recognition unit # 2 of the first scan of the second line 2L of the form 6a ends immediately before the cutout of the first character "8" of the first scan of the first line of the form 6b. The control unit 2 starts editing. Here, the editing process (1, 2, 1) is entered, the correctness / non-reading in the first scan is determined, and the character for the second scan (in this embodiment, two characters "0" and "2") ) Is selected.

The recognition unit # 1 recognizes the final character "2" of the second scan of the first line L1 of the form 6a immediately before the cutout of the character "0" of the first scan of the first line L1 of the form 6b. End with and start editing. Here, the editing process (1,
1) and 2) to determine whether the second scan is correct or unreadable.

In the case of this embodiment, the form 6a is used.
The recognition of the second scan of the first line 1L in FIG. 6 is correct reading, and therefore the recognition of the first line 1L of the form 6a is the form 6
The process ends while recognizing the first line 1L of b.

Then, the character cutting for the second scan of the second line of the form 6b is started after the character cutting for the first scan of the first line of the form 6b is completed. Form 6b
Recognition unit # 2 for the last character "1" of the first scan in the first line of
When the recognition in step 2 is finished, the editing process (2, 1, 1) is started, and it is determined whether the reading is correct or not in the first scan, and the second scanning (not shown) is performed.
Select the character to scan.

The recognition of the final character "7" of the second scan in the second line of the form 6a by the recognition unit # 1 is completed during the first scan in the second line of the form 6b, and the editing process (1,
Enter 2) and 2), and it is judged whether it is correct or not. In the case of this embodiment, since the document is correctly read, the recognition of the form 6a ends.

All the lines of the form 6a (1 in this embodiment)
Since the recognition of the lines 6 (L and 2L) is completed at this point, the image of the form 6a in the bank a of the multi-valued image memory 23 is not necessary, and the form 6a is ejected by the mechanism unit (not shown) and newly shown. The third sheet is conveyed, photoelectrically converted into a multi-valued image by the optical system 7, and stored in the bank a of the multi-valued image memory 23.

When the cutout by the control unit 2 is completed before the recognition of the recognition unit of the first line 1L or the second line 2L of the form 6b is not completed, the control unit 2 sets the bank of the multi-valued image memory 23. Repeated scanning of the form 6b is illustrated so that the character cutting of the third form (not shown) stored in a may be started and the repeated scan of the form 6a may be executed during the recognition process of the form 6b. It may be executed during the recognition processing of the third sheet, and the same processing is repeated thereafter.

Therefore, in the optical character reading apparatus according to this embodiment, the image memory 23 for storing the images of the plurality of sheets 6a and 6b as a multi-valued image is provided, and the plurality of sheets of the plurality of sheets 6a and 6b are stored immediately after the character cutting is completed. Report (6
a, 6b) character cutting out of a line is performed, and after character recognition of that line is completed, it is determined whether the line is correct or not, and if a character to be reread occurs, a line between multiple forms Since the character cut-out of the line is performed after the character cut-out is completed, the character cut-out section and the recognition section operate efficiently over the recognition processing of a plurality of forms, and the processing time can be significantly improved.

In this embodiment, the multivalued image memory 2
3 stores two forms 6a and 6b, and the recognition unit is # 1,
Although three of # 2 and # 3 are used, the multi-valued image memory 23 may store any number of forms as long as it has two or more forms, and any number of recognition units as long as it has one or more. Is.

Further, the present invention can be applied to a case where only the first scan is performed without repeating the scan.

Further, in the case of recognizing only the first scan, it is not necessary to use a multi-valued image memory, and a binary image memory is sufficient.

Further, in the present embodiment, an example of processing on a line-by-line basis has been shown, but the editing unit is arbitrary, whether on a recognition character group basis on a field basis or a recognition character group basis on a plurality of lines basis.
It is also possible to use a recognition character group for each form.

[0060]

As described above, according to the optical character reading apparatus of the present invention, the character cutting of the next line is executed immediately after the character cutting of each line, and after the character recognition of that line is completed, If unread is determined and reread occurs,
After cutting out a line for recognizing reread characters, or after cutting out a line in the same form or in a unit of a recognized character group in another form stored in the image memory, the line is cut out. Since it was done, the character cutout part and the recognition part operate efficiently,
It is expected that the processing time can be significantly improved.

[Brief description of drawings]

FIG. 1 is a block diagram of an optical character reading device shown as an embodiment of the present invention.

FIG. 2 is a diagram showing an example of a form used in the same device shown in FIG.

FIG. 3 is a time chart for explaining a character recognition process of the above-described device shown in FIG.

FIG. 4 is a block diagram of an optical character reading device shown as another embodiment of the present invention.

5 is a diagram showing an example of a form used in the same apparatus shown in FIG.

6 is a time chart for explaining a character recognition process of the same apparatus shown in FIG.

FIG. 7 is a block diagram showing a configuration example of a conventional optical character reading device.

FIG. 8 is a diagram showing an example of a form used in the same device shown in FIG. 7.

9 is a time chart for explaining a character recognition process of the same apparatus shown in FIG.

[Explanation of symbols]

 1 address data bus 2 control unit 3 multi-valued image memory 4 comparator 5 pattern memory 6 form 6a form 6b form 7 optical system 23 multi-valued image memory # 1 to # 3 recognition unit

Claims (2)

[Claims] 1. A photoelectric conversion unit that optically reads a form and converts it into a multivalued image, an image memory that stores an image of the form photoelectrically converted into a multivalued image, and a plurality of slices in the density direction from the image memory. The form image is read out by switching between the characters, and a character cutout unit that cuts out characters for each character and a recognition unit that recognizes the cutout characters by one or a plurality of recognition units are provided. An optical character reading device for performing recognition and recognizing recognition character group units such as row units or field units by switching a predetermined slice based on the recognition result. Characters are cut out for each recognized character group, and after the recognition for each recognized character group ends It is characterized in that a control means is provided for judging non-reading and, when a character requiring rereading occurs, performing character cutting out for recognition of the rereading character after cutting out a character in a certain recognized character group unit. Optical character reader. 2. A photoelectric conversion unit for optically reading a form and converting it into a multi-valued image, an image memory for storing a plurality of images of the form photoelectrically converted into a multi-valued image, and a plurality of slices from the image memory. It is composed of a character cutout unit that reads out the form image while switching between the two, and cuts out characters for each character, and a recognition unit that recognizes the cut out characters with one or more recognition units. An optical character reading device for recognizing a recognition character group unit such as a line unit or a field unit, which is recognized again by switching a predetermined slice based on the recognition result, and immediately after the character cutting of the recognition character group unit is completed, In addition to executing the character cut-out for each recognized character group in the same form or other forms stored in the image memory After the recognition of each recognition character group unit, it is determined whether the recognition character group unit is correct or non-readable, and if a reread character occurs, a character cutout for recognition of the reread character is made in the same form or in the image memory. An optical character reading device provided with a control means for cutting out characters in units of recognized character groups with other forms stored in the table.