JPH07129713A - Character recognition device - Google Patents

Abstract

PURPOSE:To provide a character recognition device capable of character recognition processing with a high precision without troubling an operator. CONSTITUTION:A binarizing part 12 which binarizes multilevel picture data stored in a multilevel picture memory 11 by plural thresholds and plural picture memories 14, 15, and 16 for storing plural binary picture data obtained by the binarizing part 12 are provided. Further, a recognition precision calculating part 23 which calculates the character recognition precision with respect to plural binary picture data stored in plural picture memories is provided. Binary picture data having the highest recognition precision out of binary picture data stored in plural picture memories 14, 15, and 16 is selected by a discriminating part 25. The result of the character recognition processing related to selected binary picture data is displayed on a display part 26.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for recognizing characters in a document image read by an image input device.

[0002]

2. Description of the Related Art FIG. 9 is a block diagram showing a functional configuration of a conventional character recognition device.

In FIG. 9, a document image of an original is read by an image scanner 10, and the read document image is given to a binarizing unit 12 as multivalued image data. Two
The binarizing unit 12 binarizes the multi-valued image data with a predetermined threshold value and supplies the obtained binary image data to the binary image memory 1. That is, the binarization unit 12
Stores "1" in the binary image memory 1 when the multi-valued image data is larger than a predetermined threshold value, and "0" when the multi-valued image data is smaller than the predetermined threshold value. 2
The value image memory 1 is stored.

The area specifying unit 18 specifies the range of the area to be recognized in the binary image data in the binary image memory 1. The character cutout unit 19 cuts out a character pattern from the image pattern in the area designated by the area designation unit 18. The feature extraction unit 20 extracts the feature of the character from the character pattern.

The character recognition unit 21 compares the character features extracted by the feature extraction unit 20 with the standard features stored in the recognition dictionary 22, and displays the character of the standard feature having the highest similarity as a recognition candidate character. It is displayed on the section 26.

[0006]

In the above-described conventional character recognition apparatus, if the threshold value is not appropriate when binarizing by the binarizing unit 12, the binarized image may be crushed or blurred. However, there is a problem that the accuracy of character recognition by the character recognition unit 21 decreases.

In this case, in order to ensure the recognition accuracy in the conventional character recognition device, the operator changes the threshold value of the binarization unit 12 while reading the document and recognizing the character by the image scanner 10. Must be repeated.

Therefore, an object of the present invention is to provide a character recognition device which can automatically recognize the characters of a document with high accuracy without bothering the operator.

[0009]

Means for Solving the Problems (1) First Invention A character recognition device according to a first invention comprises a first storage means and a second storage means.
A value conversion means, a second storage means, a recognition accuracy calculation means, and a selection means are provided.

The first storage means stores the read document image as multi-valued image data. The binarizing means is the first
The multi-valued image data stored in the storage means is binarized with a plurality of different threshold values to obtain a plurality of binary image data. The second storage means stores a plurality of binary image data obtained by the binarization means. The recognition accuracy calculation means calculates character recognition accuracy for each of the plurality of binary image data stored in the second storage means. The selection means compares the plurality of recognition accuracies obtained by the recognition accuracy calculation means, and
The binary image data having the highest recognition accuracy is selected from the plurality of binary image data stored in the storage means.
(2) Second Invention A character recognition device according to the second invention comprises a reading means, a binarization means, a storage means, a recognition accuracy calculation means and a control means.

The reading means reads the document image as multi-valued image data. The binarizing means binarizes the multi-valued image data read by the reading means with a predetermined threshold value to obtain 2
Get value image data. The storage means stores the binary image data obtained by the binarization means. The recognition accuracy calculation means calculates the character recognition accuracy for the binary image data stored in the storage means. The control means compares the recognition accuracy obtained by the recognition accuracy calculation means with a predetermined value, causes the reading means to read the document image again when the recognition accuracy is lower than the predetermined value, and causes the binarization means to set a threshold value. Is changed to perform binarization. (3) Third Invention A character recognition device according to a third invention comprises a first storage means, a threshold value determination means, a binarization means, a second storage means, a recognition accuracy calculation means and a control means. .

The first storage means stores the read document image as multi-valued image data. The binarizing means binarizes the multi-valued image data stored in the first storage means with the threshold value determined by the threshold value determining means to obtain binary image data. The second storage means stores the binary image data obtained by the binarization means. The recognition accuracy calculation means calculates character recognition accuracy for the binary image data stored in the second storage means. The control means compares the recognition accuracy obtained by the recognition accuracy calculation means with a predetermined value, and when the recognition accuracy is lower than the predetermined value, causes the threshold value determination means to change the threshold value and changes to the binarization means. The binarization is performed with the set threshold value.

[0013]

(1) First invention In the character recognition apparatus according to the first invention, multi-valued image data is binarized with a plurality of different threshold values, and the obtained plurality of binary image data are respectively processed. Character recognition accuracy is calculated. Then, the binary image data having the maximum recognition accuracy is selected from the plurality of binary image data. Thereby, the binary image data with the highest recognition accuracy is automatically obtained. (2) Second Invention In the character recognition device according to the second invention, reading of a document image and binarization of multi-valued image data are repeated while automatically changing the threshold value. Thereby, binary image data with high recognition accuracy is automatically obtained. (3)
Third Invention In the character recognition device according to the third invention, binarization of multi-valued image data is repeated while automatically changing the threshold value. Thereby, binary image data with high recognition accuracy is automatically obtained.

[0014]

FIG. 1 is a block diagram showing the functional arrangement of a character recognition apparatus according to the first embodiment of the present invention.

In this character recognition device, a multi-valued image memory 11 for storing multi-valued image data is provided between the image scanner 10 and the binarization unit 12. The binarization unit 12 of this embodiment binarizes the multi-valued image data with three types of threshold values to obtain three binary image data. The binary image memory 13 includes a first image memory 14, a second image memory 15 and a third image memory 16 corresponding to the three binary image data. First image memory 14, second image memory 15
And the third image memory 16 stores the three binary image data obtained by the binarization unit 12, respectively.

Between the binary image memory 13 and the area designating section 18, an image for selecting one of the binary image data in the first image memory 14, the second image memory 15 and the third image memory 16 is selected. A selection unit 17 is provided.

The functions of the area designating unit 18, the character cutting unit 19, the feature extracting unit 20, the character recognizing unit 21, and the recognition dictionary 22 are the same as those of the corresponding parts in the conventional character recognizing device shown in FIG. .

Between the character recognition section 21 and the display section 26,
A recognition accuracy calculation unit 23 that calculates the accuracy of the recognition result obtained by the character recognition unit 21, a storage processing unit 24 that controls storage processing, and a determination unit 25 that makes various determinations are provided.
The determination unit 25 constitutes a selection unit.

FIG. 2 is a block diagram showing the hardware configuration of the character recognition device according to the first embodiment. 1 and 2, the same reference numerals indicate the same or corresponding parts.

This character recognition device is provided with an image scanner 1.
0, the central processing unit 100, the display unit 26, the keyboard 101, the read only memory 102 and the random access memory 105.

The read-only memory 102 stores the recognition dictionary 2
2. It includes a program storage area 103 and a threshold storage area 104. The recognition dictionary 22 stores standard features of various characters. In the program storage area 103, FIG.
A program for executing the function of each block shown in is stored. In the threshold storage area 104, 2 in FIG.
A plurality of threshold values used for binarization by the binarization unit 12 are stored. This threshold value is a reference value when classifying pixels having three or more values into white pixels or black pixels.

The random access memory 105 includes a binary image memory 13, a first character code area 106, a second character code area 107, a third character code area 108, a final character code area 109, an image memory instruction flag 110, and a multivalued. Image memory 11, work area 111, first recognition accuracy area 1
12, a second recognition accuracy region 113 and a third recognition accuracy region 114 are included.

The binary image memory 13 corresponds to the binary image memory 13 shown in FIG. The first character code area 106,
Second character code area 107 and third character code area 1
In 08, the character codes of the recognition candidate characters obtained by the character recognition regarding the binary image data in the first image memory 14, the second image memory 15, and the third image memory 16, respectively are stored.

In the final character code area 109, among the character codes stored in the first character code area 106, the second character code area 107 and the third character code area 108, the character code obtained with the highest recognition accuracy. Is memorized. The multi-valued image memory 11 corresponds to the multi-valued image memory 11 shown in FIG.

The first recognition accuracy area 112 stores the recognition accuracy in character recognition regarding the binary image data in the first image memory 14. In the second recognition accuracy area 113,
The recognition accuracy in character recognition regarding the binary image data in the second image memory 15 is stored. Third recognition accuracy area 1
The recognition accuracy in character recognition regarding the binary image data in the third image memory 16 is stored in 14.

The image memory instruction flag 110 is one of binary image data stored in the first image memory 14, the second image memory 15 and the third image memory 16, and should be input to the character cutout unit 19 of FIG. Value image data is shown. Specifically, the image memory instruction flag 110 indicates that the binary image data in the first image memory 14, the binary image data in the second image memory 15, and the binary image in the third image memory 16 are used in the character recognition process. When specifying data, they are set to "-1,""-2," and "-3," respectively.

The image memory instruction flag 110 is set to "1" when it is determined that the character recognition accuracy of the binary image data in the first image memory 14 is the highest, and the image memory instruction flag 110 in the second image memory 15 is set. "2" when it is found that the recognition accuracy of character recognition for binary image data of
And is set to "3" when it is found that the recognition accuracy of the character recognition regarding the binary image data in the third image memory 16 is the highest.

Next, the operation of the character recognition apparatus shown in FIGS. 1 and 2 will be described with reference to the flowchart of FIG.

First, in step S1, the document image is read from the original by the image scanner 10 and stored in the multivalued image memory 11 as multivalued image data. In step S2, the binarization unit 12 causes the threshold storage area 104
Converting the multi-valued image data in the multi-valued image memory 11 into three binary image data using the three types of threshold values stored in the first image memory 14;
It is stored in the second image memory 15 and the third image memory 16, respectively.

For example, when the multi-valued image data has 256 gradations, three appropriate threshold values of "64", "128" and "192" are predetermined. In the first image memory 14, "1" is stored for pixels having a density of one pixel larger than "64", and "0" is stored for pixels having a density of one pixel smaller than "64". Similarly, in the second image memory 15, "1" is stored for a pixel whose density of one pixel is higher than "128", and "0" is stored for a pixel whose density of one pixel is lower than "128". In the third image memory 16, “1” is set for pixels in which the density of one pixel is larger than “192”.
Is stored, and “0” is stored for a pixel in which the density of one pixel is smaller than “192”.

Then, in step S3, the image memory instruction flag 110 is set to "-1". Step S4
Then, the image selection unit 17 causes the image memory instruction flag 110 to be displayed.
Based on the first image memory 14 and the second image memory 15
Then, of the binary image data stored in the third image memory 16, which binary image data is to be subjected to the character recognition process is selected. Further, the area designating section 18 designates a recognition target area in the binary image data in the image memory selected by the image selecting section 17. For example, in the binary image data stored in the first image memory 14, it is possible to specify a rectangular recognition target area whose diagonal is a straight line connecting the point (10, 120) and the point (100, 330). Details of the process of step S4 will be described later with reference to the flowchart of FIG.

In step S5, the character cutting section 19
A character pattern is cut out from the recognition target area designated by the area designating unit 18. In step S6, the feature extraction unit 2
0 extracts the characteristics of each character with respect to a predetermined number of character patterns cut out by the character cutting unit 19. In step S7, the character recognition unit 2
1 compares the feature of each character extracted by the feature extraction unit 20 with the standard feature stored in the recognition dictionary 22, and determines the degree of similarity. Then, the standard feature character having the maximum similarity is determined as the recognition candidate character.

In step S8, the recognition accuracy calculation unit 23 calculates the recognition accuracy of the character recognition by the character recognition unit 21.
Generally, in character recognition, when the maximum similarity is larger than a predetermined value or when the difference between the maximum similarity and the next largest similarity (hereinafter referred to as similarity difference) is larger than a predetermined value, the recognition candidate character is The answer is correct, and when the maximum similarity and the difference in similarity are smaller than predetermined values, the recognition candidate character tends to be incorrect. So for each character,
When the maximum similarity or similarity difference is larger than a predetermined value, the character can be "determined", and when the maximum similarity or similarity difference is smaller than a predetermined value, the character cannot be "determined". Assuming that there is, the number of confirmed characters and the number of unconfirmed characters are calculated. Then, the recognition accuracy is defined by the number of confirmed characters / (the number of confirmed characters + the number of unconfirmed characters).

In step S9, the storage processing unit 24 stores the character code of the recognition candidate character obtained in step S7 in the first character code area 106, the second character code area 107 or the third character code area 108, Step S
The recognition accuracy obtained in 8 is stored in the first recognition accuracy area 112, the second recognition accuracy area 113, or the third recognition accuracy area 114. For details of the processing in step S9, see FIG.
It will be described later with reference to the flowchart of.

Next, in step S10, the determination unit 25
Determines whether the image memory instruction flag 110 is "-3". The image memory instruction flag 110 is "-
When it is "3", the process proceeds to step S12, and when the image memory instruction flag 110 is "-1" or "-2", the process proceeds to step S11.
When it is "1" or "-2", three binary image memories 1
This is a state in which the calculation of the recognition accuracy has not been completed for any of 4, 15, and 16.

In step S11, the determination unit 25 subtracts 1 from the value of the image memory instruction flag 110, and then the process returns to step S4 and the image memory instruction flag 110 is "-3".
The processing of steps S4 to S9 is repeated until.

In step S12, the determination section 25 determines the first image memory 14, the second image memory 15, and the third image memory 1 based on the recognition accuracy obtained in step S8.
The image memory for storing the binary image data having the highest recognition accuracy out of 6 is selected, and the corresponding value among “1”, “2” and “3” is set in the image memory instruction flag 110.

In step S13, the storage processing unit 24
The character code in the character code area corresponding to the image memory selected in step S12 is set to the final character code area 109.
To memorize. That is, when the first image memory 14 is selected, the character code in the first character code area 106 is copied to the final character code area 109, and when the second image memory 15 is selected, the second character is selected. Code area 10
The character code in 7 is copied to the final character code area 109, and when the third image memory 16 is selected, the character code in the third character code area 108 is copied to the final character code area 109.

In step S14, the feature extraction unit 20
In the binary image data in the image memory selected in step S12, character features are extracted from the remaining character patterns not extracted in step S6. Step S1
In step 5, the character cutout unit 19 executes step S21 described later.
The character pattern is cut out from the recognition target area specified by. In step S16, the feature extraction unit 20 extracts the feature of the character from the character pattern cut out in step S15.

In step S17, the character recognition unit 21
Comparing the character features extracted in step S14 or step S16 with the standard features stored in the recognition dictionary 22,
Determine the degree of similarity. Then, the character code of the standard feature character having the maximum similarity is determined. In step S18, the storage processing unit 24 stores the character code obtained in step S17 in the final character code area 109. In step S19, the determination unit 25 causes the display unit 26 to display the character code stored in the final character code area 109.

In step S20, it is determined whether or not character recognition processing is to be performed on another area. If so, the process proceeds to step S21, and if not, the processing ends. In step S21, the recognition target area is specified by the area specifying unit 18, and the process proceeds to step S15.

Next, referring to the flow chart of FIG.
The processing of the image selection unit 17 in step S4 of will be described.

First, the image memory instruction flag 110 is set to "-.
It is determined whether or not it is 1 "(step S100). When the image memory instruction flag 110 is" -1 ", the first image memory 14 is selected as a target of the character recognition process (step S103). The area designation unit 18 designates a recognition target area in the binary image data in the first image memory 14 (step S104).

When the image memory instruction flag 110 is not "-1" in step S100, it is determined whether or not the image memory instruction flag 110 is "-2" (step S101). When the image memory instruction flag 110 is "-2", the second image memory 1 is selected as the target of the character recognition process.
5 is selected (step S105). In this case, the character recognition processing is performed on the recognition target area designated when the first image memory 14 is selected.

When the image memory instruction flag 110 is not "-2" in step S101, it is determined whether the image memory instruction flag 110 is "-3" (step S102). When the image memory instruction flag 110 is "-3", the third image memory 1 is selected as the target of the character recognition processing.
6 is selected (step S106). Also in this case, the character recognition process is performed on the recognition target area designated when the first image memory 14 is selected.

If the image memory instruction flag 110 is not "-3" in step S102, the process is terminated.

Next, referring to the flow chart of FIG.
The processing of the storage processing unit 24 in step S9 will be described.

First, the image memory instruction flag 110 is set to "-".
It is determined whether it is 1 "(step S200). When the image memory instruction flag 110 is" -1 ", the process proceeds to step S204. In step S204, the binary image data in the first image memory 14 is processed. This is a process performed after performing the character recognition process.
The character code of the recognition candidate character is stored in the first character code area 106, and in step S205, the recognition accuracy of the binary image data in the first image memory 14 is stored in the first recognition accuracy storage unit 112.

When the image memory instruction flag 110 is not "-1" in step S200, it is determined whether the image memory instruction flag 110 is "-2" (step S201). When the image memory instruction flag 110 is "-2", the process proceeds to step S206. Step S206
Is a process performed after the character recognition process is performed on the binary image data in the second image memory 15. In step S206, the character code of the recognition candidate character is stored in the second character code area 107, and in step S207, the recognition accuracy regarding the binary image data in the second image memory 15 is stored in the second recognition accuracy storage unit 113. .

When the image memory instruction flag 110 is not "-2" in step S201, the process proceeds to step S202. Step S202 is a process performed after the character recognition process is performed on the binary image data in the third image memory 16. In step S202, the character code of the recognition candidate character is stored in the third character code area 108, and in step S203, the character code of the recognition candidate character is stored in the third recognition accuracy storage unit 114.
The recognition accuracy regarding the binary image data in the image memory 16 is stored.

Here, the operation of the character recognition apparatus of the first embodiment will be described using a specific example. FIG. 6A is a diagram schematically showing the layout of a document read by the image scanner 10. In practice, for example, FIG.
Character patterns and other image patterns are printed as shown in FIG.

The original shown in FIG. 6A is the image scanner 1.
0 is read and stored in the multivalued image memory 11 as multivalued image data. After that, the multi-valued image data in the multi-valued image memory 11 is converted into three binary image data by the binarization unit 12 using three kinds of threshold values, and the first image memory 14 and the second image memory are respectively converted. 15 and the third image memory 16 are stored.

Here, it is assumed that the operator has designated the area 1 in FIG. 6A as the recognition target area by the area designating unit 18. It is assumed that the character pattern shown in FIG. 6B is the actual character pattern of the area 1. One of the character patterns cut out by the feature extraction unit 20 after the character pattern included in the area 1 is cut out by the character cutting unit 19
Character features are extracted for 0 characters, and the character recognition unit 2
In 1, the character recognition processing is performed.

FIG. 6C shows a state in which the recognition candidate characters are stored in the first character code area 106, the second character code area 107 and the third character code area 108. In the figure, the characters surrounded by a rectangle in the "contents of storage location" column are the characters that can be confirmed because the maximum similarity or the difference in similarity is higher than a predetermined threshold value. Characters that are not enclosed in a rectangle are characters that could not be determined.

In this case, the recognition accuracy obtained from the number of confirmed characters and the number of unconfirmed characters is as shown in FIG. 6 (d).
In this example, it can be seen that the recognition accuracy is highest when characters are recognized by the second image memory 15. Therefore, in the subsequent processing, the character recognition processing is performed on the binary image data in the second image memory 15 even for the 11th and subsequent character patterns shown in FIG. 6B. In addition, the character recognition processing is also performed on the binary image data in the second image memory 15 in areas other than the area 1.

As described above, according to the character recognition apparatus of the first embodiment, multi-valued image data is converted into a plurality of binary image data using a plurality of threshold values, and a binary image with high recognition accuracy is obtained. Since the character recognition processing is performed using the data, the character recognition processing is performed with high accuracy by reading the original once.

In the above embodiment, the binarization is performed with three thresholds, but any number of thresholds may be used as long as it is two or more.

FIG. 7 is a block diagram showing the functional arrangement of a character recognition device according to the second embodiment of the present invention.

The character recognition device of FIG. 7 differs from the character recognition device of FIG. 1 in that the multivalued image memory 11 and the image selection unit 1 are used.
7 is not provided, and it is composed of 3 image memories 2
A point that a binary image memory 1 composed of one image memory is provided in place of the value image memory 13, and the determination unit 2
5 is the difference. The determination unit 25 constitutes a control means.

First, the original is read by the image scanner 10 as multivalued image data. The binarization unit 12
The multi-valued image data output from the image scanner 10 is binarized with a predetermined threshold value, and the binary image data is stored in the binary image memory 1. The operations of the area designation unit 18, the character cutout unit 19, the feature extraction unit 20, the character recognition unit 21, and the recognition accuracy calculation unit 23 are the same as those in the first embodiment. The recognition accuracy calculation unit 23 calculates the character recognition accuracy of the binary image data in the binary image memory 1, and the storage processing unit 24
Are stored in the random access memory.

The judging section 25 judges whether or not the recognition accuracy is higher than a predetermined value. When the recognition accuracy is lower than the predetermined value, the determination unit 25 causes the image scanner 10 to read the original again as multivalued image data. The binarization unit 12 binarizes the multi-valued image data output from the image scanner 10 with a threshold value different from the previous threshold value and stores the binary image data in the binary image memory 1. afterwards,
The same processing as the previous time is performed, and the recognition accuracy calculation unit 23 calculates the character recognition accuracy. Similar processing is repeated until the recognition accuracy becomes higher than a predetermined value or the binarization processing exceeds a predetermined number of times.

As described above, according to the character recognition apparatus of the second embodiment, the binarization process is repeated while automatically reading the original and changing the threshold value. As with the character recognition device, character recognition processing is performed with high accuracy.

FIG. 8 is a block diagram showing the functional arrangement of a character recognition device according to the third embodiment of the present invention.

The character recognition device of FIG. 8 differs from the character recognition device of FIG. 1 in that a binarization threshold value determination unit 30 is newly provided in the binary image memory 13 including three image memories. Instead, the binary image memory 1 including one image memory is provided, the image selection unit 17 is not provided, and the processing of the determination unit 25 is different. The determination unit 25 constitutes a control means.

First, the document is read as multi-valued image data by the image scanner 10, and the multi-valued image data is stored in the multi-valued image memory 11. The binarization threshold value determination unit 30 determines a threshold value for binarization. The binarization unit 12 binarizes the multivalued image data in the multivalued image memory 11 using the determined threshold value and stores the binary image data in the binary image memory 1. The operations of the area designation unit 18, the character cutout unit 19, the feature extraction unit 20, the character recognition unit 21, and the recognition accuracy calculation unit 23 are the same as those in the first embodiment. The recognition accuracy calculation unit 23 causes the binary image memory 1
The recognition accuracy of the character relating to the value image data is calculated and stored in the random access memory by the storage processing unit 24.

The judging section 25 judges whether or not the recognition accuracy is higher than a predetermined value. When the recognition accuracy is lower than the predetermined value, the determination unit 25 determines the binarization threshold value determination unit 30.
To change the threshold. The binarization unit 12 sets the multi-valued image data in the multi-valued image memory 11 to 2 with the changed threshold value.
The binary image data is binarized and stored in the binary image memory 1. Thereafter, the same processing as the previous time is performed, and the recognition accuracy calculation unit 23 calculates the character recognition accuracy. Similar processing is repeated until the recognition accuracy becomes higher than a predetermined value or the binarization processing exceeds a predetermined number of times.

As described above, according to the character recognition apparatus of the third embodiment, since the binarization processing is repeated while automatically changing the threshold value, it is the same as the character recognition apparatus of the first embodiment. In addition, character recognition processing is performed with high accuracy.

[0068]

As described above, according to the first, second and third inventions, since the binary image data with high recognition accuracy is automatically obtained, the characters of the document image can be displayed without bothering the operator. There is provided a character recognition device capable of recognizing characters with high accuracy.

[Brief description of drawings]

FIG. 1 is a block diagram showing a functional configuration of a character recognition device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a hardware configuration of a character recognition device according to the first embodiment of the present invention.

FIG. 3 is a flowchart of the operation of the character recognition device in the first embodiment of the present invention.

FIG. 4 is a flowchart showing a processing procedure of an image selection unit in FIG.

5 is a flowchart showing a processing procedure of a storage processing unit in FIG.

FIG. 6 is a diagram for explaining a specific operation of the character recognition device according to the first embodiment of the present invention.

FIG. 7 is a block diagram showing a functional configuration of a character recognition device according to a second embodiment of the present invention.

FIG. 8 is a block diagram showing a functional configuration of a character recognition device according to a third embodiment of the present invention.

FIG. 9 is a block diagram showing a functional configuration of a conventional character recognition device.

[Explanation of symbols]

 10, 100 image scanner 11 multi-valued image memory 12 binarization unit 13 binary image memory 14 first image memory 15 second image memory 16 third image memory 17 image selection unit 18 area designation unit 19 character cutout unit 20 feature extraction Part 21 Character recognition part 22 Recognition dictionary 23 Recognition accuracy calculation part 24 Storage processing part 25 Judgment part 26 Display part 30 Binarization threshold value determination part 101 Keyboard 102 Read only memory 103 Program storage area 104 Threshold storage area 105 Random access Memory 106 First character code area 107 Second character code area 108 Third character code area 109 Final character code area 110 Image memory instruction flag 111 Work area 112 First recognition accuracy area 113 Second recognition accuracy area 114 Third recognition accuracy area

Claims (3)

[Claims] 1. A first storage unit for storing a read document image as multivalued image data, and binarizing the multivalued image data stored in the first storage unit with a plurality of different threshold values. Binarizing means for obtaining a plurality of binary image data, a second storing means for storing the plurality of binary image data obtained by the binarizing means, and a second storing means The recognition accuracy calculation means for calculating the recognition accuracy of a character with respect to a plurality of binary image data and the plurality of recognition accuracies obtained by the recognition accuracy calculation means are compared with each other, and a plurality of two recognition values stored in the second storage means are compared. A character recognition device, comprising: selection means for selecting the binary image data having the highest recognition accuracy among the value image data. 2. A reading means for reading a document image as multivalued image data, and a binarizing means for binarizing the multivalued image data read by said reading means with a predetermined threshold value to obtain binary image data. A storage unit for storing the binary image data obtained by the binarization unit; a recognition accuracy calculation unit for calculating character recognition accuracy for the binary image data stored in the storage unit; The recognition accuracy obtained by the calculation means is compared with a predetermined value, and when the recognition accuracy is lower than the predetermined value, the reading means is made to read the document image again and the binarization means is caused to perform the predetermined operation. A character recognition device, comprising: a control unit that changes a threshold value to perform binarization. 3. A first storage means for storing the read document image as multi-valued image data, a threshold value determination means for determining a threshold value for binarization, and the first storage means. Binarizing means for binarizing the multivalued image data stored in the binarized image by the threshold value determined by the threshold value determining means to obtain binary image data; and the binary value obtained by the binarizing means. Second storage means for storing image data, recognition accuracy calculation means for calculating character recognition accuracy for binary image data stored in the second storage means, and recognition obtained by the recognition accuracy calculation means The accuracy is compared with a predetermined value, and when the recognition accuracy is lower than the predetermined value, the threshold value determining means changes the threshold value and the binarizing means changes the binary value with the threshold value. And a control means for performing Character recognition device.