JP2000306042A - Device and method for recognizing character - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device for recognizing character in which a template storage area is decreased, matching processing time is reduced and a special matching mechanism is unnecessitated and to provide a method therefor. SOLUTION: This device is provided with a template storage part 15 for storing a reduced template obtained by reducing a standard template, a restoration information storage part 16 for storing restoration information for restoring the reduced template to the standard template, an input pattern reducing part 21, a standard template restoring part 22 and a matching processing part 23 for matching the reduced pattern and the reduced template and matching the input pattern and the standard template later. Since the standard template is restored on the basis of the restoration information for restoring the reduced template to the standard template, data can be compressed and the template storage area can be decreased. Further, since matching is performed while using the reduced template, matching processing time can be reduced.

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 and a character recognition method thereof, and more particularly to a character recognition device and a character recognition method suitable for performing character recognition by a template matching method.

[0002]

2. Description of the Related Art As a conventional character recognition method, there is a template matching method. This template matching method is a method in which an input pattern to be subjected to character recognition and a pre-registered template are sequentially superimposed to obtain the most matching template, and the image data is converted into character character data.

[0003] As a conventional example of the above character recognition,
For example, a technique described in Japanese Patent Application Laid-Open No. Sho 60-225986 has been proposed. This gazette aims to process simple characters at high speed based on the global characteristics of input characters, and to process slips at high speed in the case of data entry such as alphanumeric characters and in the case of kanji mixed Japanese input. Means for extracting global features of the input character, such as the complexity of the input character pattern,
There is disclosed a character recognition device characterized by means for normalizing the size of an input character pattern by a value obtained by this means, and performing identification processing using the normalized pattern obtained by said means. I have.

[0004]

However, the above-described prior art has the following problems.

[0005] In the conventional method of sequentially overlapping an input pattern to be recognized and a pre-registered template to obtain a template with the best match, the template that must be registered increases as the number of character types to be recognized increases. Therefore, there is a problem that an area for storing a template increases due to an increase in the number of character types. Further, since the superposition with the template is sequentially performed, there is a problem that the processing time increases with an increase in the template.

SUMMARY OF THE INVENTION An object of the present invention is to provide a character recognition device and a character recognition method which reduce the template storage area and the matching processing time, and which do not require a special matching mechanism.

[0007]

SUMMARY OF THE INVENTION The present invention relates to a character recognition apparatus for performing character recognition based on template matching, wherein a reduced template storage means for storing a reduced template obtained by reducing a standard template, and a reduced input pattern to be recognized. A reducing unit that creates a pattern, and a matching processing unit that performs matching between the reduced pattern created by the reducing unit and the reduced template, and then performs matching between the input pattern and the standard template. Features.

The present invention also relates to a character recognition method for performing character recognition based on template matching, wherein a character is cut out and a size of each character is normalized from an image to be subjected to the character recognition, and the cut and normalization are performed. A reduced pattern is created based on the reduced pattern, the matching between the reduced pattern and a reduced template obtained by reducing the standard template is performed first, and then the matching between the cut-out and normalized pattern and the standard template is performed. It is characterized by the following.

A character recognition apparatus according to the present invention, which will be described with reference to FIG. 2, is a character recognition apparatus for performing character recognition based on template matching. The reduced template storage means stores a reduced template obtained by reducing a standard template. (15 in FIG. 2) and reduction means (2 in FIG. 2) for generating a reduction pattern from an input pattern to be subjected to character recognition.
1) and matching processing means (23 in FIG. 2) for matching the reduced pattern created by the reducing means with the reduced template and then matching the input pattern with the standard template. ing.

[Operation] The character recognition apparatus of the present invention stores a reduced template obtained by reducing the standard template, and restores the standard template based on restoration information for restoring the reduced template to the standard template. For this reason, data compression can be realized, and the template storage area can be reduced. Further, matching is performed using a reduced template. For this reason, the matching processing time can be reduced. Further, the standard template is restored from the reduced template and the restoration information to perform matching. Therefore, the effect can be improved without lowering the recognition performance.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment Next, a first embodiment of the present invention will be described in detail with reference to the drawings.

(1) Description of Configuration FIG. 1 is a block diagram showing an overall configuration example of a character recognition device according to a first embodiment of the present invention. 1, the character recognition apparatus according to the first embodiment of the present invention includes a scanner unit 11, a preprocessing unit 12, a matching unit 13, a determination unit 14, a template storage unit 15, and a restoration information storage unit 16.

The above configuration will be described in detail.
Reads a character recognition target image from a document as image data using a photoelectric conversion element or the like, and outputs the image data to the preprocessing unit 12.
The preprocessing unit 12 cuts out the character recognition target image for each character and normalizes the size, and outputs a preprocessing result to the matching unit 13. The matching unit 13 performs a matching process as described later on the preprocessing result. Judgment unit 14
Performs character recognition based on a matching process and outputs a recognition result. The template storage unit 15 stores a reduced template. The restoration information storage unit 16 stores restoration information.

In the above configuration, the recognition target image input from the scanner unit 11 is cut out for each character by the pre-processing unit 12 and the size is normalized, and then the matching unit 1
3, a matching process is performed, and the determination unit 14 outputs a recognition result.

FIG. 2 is a block diagram showing a detailed configuration example centering on the matching unit 13 of the character recognition device according to the first embodiment of the present invention shown in FIG. In FIG. 2, the matching unit 13 of the character recognition device according to the first embodiment of the present invention
Are the input pattern reduction unit 21 and the standard template restoration unit 2
2, a matching processing unit 23 and a large classification result storage unit 24 are provided.

More specifically, the input pattern reduction section 21 creates a reduced pattern from the normalized pattern input from the preprocessing section 12. The standard template restoring unit 22 creates a standard template from the reduced template 42 input from the template storage unit 15 and the restoration information 43 input from the restoration information storage unit 16.

The matching processing unit 23 stores the reduced pattern input from the input pattern reducing unit 21 and the template storage unit 1.
5, matching with the reduced template 42 input from
Alternatively, matching processing between the input pattern input from the preprocessing unit 12 and the standard template input from the standard template restoring unit 22 is performed. The former matching process with a reduced template is called a large classification matching process, and the latter matching process with a standard template is called a detailed matching process.

The large classification result storage section 24 is a buffer for temporarily storing the result of the large classification matching processing, and provides information for selecting a processing target template to the matching processing section 23 during the detailed matching processing. In FIG. 2, the flow of the large classification matching process is indicated by a solid line, and the flow of the detailed matching process is indicated by a broken line.

The function of the matching unit 13 will be described in further detail. An image for one character to be recognized is converted from a preprocessing unit 12 into a binary image normalized to a fixed size of 2n × 2n. 13 is input. The input pattern reduction unit 21 of the matching unit 13 performs linear 1/4 reduction processing on an input pattern, which is a binary image having a size of 2n × 2n, input from the preprocessing unit 12, and has a size of n × A reduced pattern which is a binary image of n is created. As a reduction method, for example, when there are two or more black pixels in four pixels of a 2 × 2 area of the input pattern, the pixels of the reduction pattern are set to black, and when the number of black pixels is less than two, the pixels of the reduction pattern are set to black. All you have to do is white.

The standard template restoring unit 22 receives a black reduced template and a white reduced template 42, each of which is a binary image of size n × n, input from the template storage unit 15, and 5 input from the restoration information storage unit 16. × n
× 2 bits from restoration information 43 consisting of n bits
A black standard template and a white standard template, which are nx2n binary images, are created. The detailed configuration and the restoring method of the standard template restoring unit will be described later.

The matching processing section 23 performs a matching process between the input image and the black and white templates. The black template indicates pixels that expect black in the input image, the white template indicates pixels that expect white in the input image, and the result of the matching process is the result of counting mismatches with these expectations. The matching processing unit 23 performs two types of matching processing as described above. The first is a large classification matching process between the reduced pattern created by the input pattern reducing unit 21 and the black reduced template and the white reduced template input from the template storage unit 15.
The third is a detailed matching process between the input pattern input from the preprocessing unit 12 and the black standard template and the white standard template created by the standard template restoring unit 22.

The detailed matching process is performed with a size of 2n × 2n, while the large classification matching process is performed with a size of n.
× n. The large classification matching process sequentially performs a matching process with all templates stored in the template storage unit 15 in advance, and outputs the ID of the template with a higher matching result to the large classification result storage unit 24. The detailed matching process sequentially performs the matching process only on the template selected by the template ID stored in the large classification result storage unit 24, and outputs the result to the determination unit 14.

FIG. 3 is a diagram showing features of the standard template according to the first embodiment of the present invention. Each standard template is configured as a set of a black template 31 and a white template 32. A black template 31 indicates a pixel expected to be black, and a white template 32
Indicates a pixel expected to be white. If the number of mismatched pixels is counted as a matching process and the smaller one is output as a recognition result, the {black template value, white template value} of one pixel having the standard template is {0, 0}, {0, There are only three states of {1} and {1,0}, and the state of {1,1} is not required.

In the black standard template 31, "1" is set to a pixel expected to be black, and in the white standard template 32, "1" is set to a pixel expected to be white.
Is set. In FIG. 3, the pixels to which '1' is set are shown in black. Since the number of mismatched pixels is counted as a matching process and the smaller one is output as a recognition result, that is, when the input of a pixel expected to be black is white and when the input of a pixel expected to be black is white Since the case where the input of the expected pixel is black is evaluated as a mismatch, the {black standard template value, white standard template value} of one pixel having the standard template does not match whether the input is black or white. {1,1}
It is not necessary.

FIG. 4 is a diagram showing characteristics of reduction and restoration in the first embodiment of the present invention. In the reduction, each of the vertical and horizontal directions is reduced to １ ／. That is, the adjacent 2 of the standard image 41
Four pixels of × 2 correspond to one pixel of the reduced image 42. According to this reduction method, rough two-dimensional characteristics are maintained, so that matching processing between reduced images can be realized at high speed and easily.

The restoration information 43 is data necessary for restoring the reduced template 42 to the standard template 41. Considering the features of the standard template 41 described above, in order to restore the standard template 41 of 2 × 2 pixels, it is only necessary to have a reduced template 42 of one pixel and restoration information 43 of 5 bits. That is, the standard template 41 of 8 bits including the black template and the white template
In order to restore the data, a total of 7 bits of the 2-bit reduced template 42 and the 5-bit restoration information 43 is sufficient. As a result, the storage area obtained by combining the template storage area and the restoration information storage area can be reduced to 7/8 as compared with the case where the standard template 41 is stored.

The reduction is the same as the method shown in the example of the reduction method of the input pattern reduction unit 21. That is, each size 2
The black standard template and the white standard template, which are n × 2n binary images, are linearly reduced by a factor of 4 to obtain a black reduced template and a white reduced template that are binary images of size n × n. Specifically, for example, the standard pattern 41
The pixel of the reduced template 42 is set to black when there are two or more black pixels in the four pixels of the 2 × 2 area, and the pixel of the reduced template 42 is set to white when there are less than two black pixels. The restoration is performed by restoring the standard pattern of 8 bits of each of 4 pixels of black and white from only the 2 bits of the reduced template of 1 pixel of each of black and white and the restoration information 43 of 5 bits.

FIG. 5 is a diagram showing an example of creating restoration information in the first embodiment of the present invention. From the characteristics of the standard template described above, the state of the adjacent four pixels of the standard template is 7 bits in total, and there are only 81 possible states shown in the table. Therefore, it is only necessary to be able to identify these 81 states. When classified by the state of the reduced template,
State of one pixel of reduced template ｛black reduced template,
White reduction template｝ is {0,0｝ ｛0,1｝ ｛1,
When 0 {1,1}, 21 ways, 27 ways, 2 ways respectively
There are 7 ways and 6 ways. Since it is sufficient that each of these can be determined by the restoration information, it is sufficient to prepare 5 bits as the restoration information. For example, data that sequentially counts up as shown in FIG. 5 may be used as the restoration information. In addition,
If the restoration information is created in this way, any creation method may be used.

(2) Description of Operation Next, the operation of the first embodiment of the present invention will be described with reference to FIGS.
This will be described in detail with reference to FIGS.

FIG. 6 is a diagram showing a flow of the operation of the matching unit 13 of the character recognition device according to the first embodiment of the present invention. The matching unit 13 first performs a large classification matching process 51. In the large classification matching process 51, the reduced templates 42 are sequentially input from the template storage unit 15, and matching is performed with the reduced patterns created by the input pattern reducing unit 21. In the example shown in FIG. 6, N reduced templates with pattern IDs of “0”, “1”, 2 ”...“ N ”are input from the template storage unit 15, and the large classification matching process 51 is performed in the order of (52). Is going.

When the matching with all the reduced templates is completed, the upper template ID is output to the large classification result storage unit 24 as a result, that is, a large classification matching result.
In the example shown in FIG. 6, three template IDs “A”, “B”, and “C” are output (53).

Next, the matching section 13 performs a detailed matching process 54. In the detailed matching process 54, template IDs are sequentially input from the large classification result storage unit 24, and the reduced template 42 and the restoration information 43 corresponding to the template ID are stored in the template storage unit 15 and the restoration information storage unit 1.
While inputting from 6 and restoring to the standard template 41,
Performs matching with the input pattern. In the example shown in FIG.
First, the template ID “A” from the large classification result storage unit 24
Is input (55), and the template ID is read from the template storage unit 15 and the restoration information storage unit 16 accordingly.
The reduced template 42 of 'A' and the restoration information 43 are input (56), and matching is performed while being restored to the standard template 41.

When the detailed matching processing of the template ID 'A' is completed, the template ID 'B'C'
Detailed matching processing is performed on a total of three templates. In the first embodiment, the large classification result storage unit 2
Although the description has been made on the assumption that the number of IDs stored in 4 is three, the number is not limited.

FIG. 7 is an explanatory diagram showing an example in which the reduced template according to the first embodiment of the present invention retains the two-dimensional characteristics of the standard template. Since the reduced template retains the two-dimensional characteristics of the standard template, matching processing can be performed in the same manner as detailed matching processing.
A special matching mechanism for the large classification matching process is not required.

According to the first embodiment of the present invention, since the standard template 41 is reduced and the reduced template 42 is stored, the template storage area can be reduced. As shown in FIG. 4, even if the restoration information 43 for restoring the standard template 41 is added, the whole can be reduced to 7/8. A detailed matching process 54 based on the result of the large classification matching process 51 which can be executed at high speed
Is limited, the processing time of the matching unit 13 can be reduced. In addition, since the reduced template 42 retains the two-dimensional characteristics of the standard template 41, the matching process can be performed in the same manner as the detailed matching process, so that a special matching mechanism for the large classification matching process is not required.

[Second Embodiment] Next, a second embodiment of the present invention will be described in detail with reference to the drawings.

(1) Description of Configuration FIG. 8 is a block diagram showing a detailed configuration example centered on the matching unit 13 of the character recognition device according to the second embodiment of the present invention. 8, the matching unit 13 of the character recognition device according to the second embodiment of the present invention includes an input pattern reduction unit 21, a matching processing unit 23, and a large classification result storage unit 24.

In the second embodiment of the present invention, a standard template storage section 71 and a template storage section (reduced template storage section) 15 are provided as template storage sections, and the restoration information storage section 16 is provided as in the first embodiment. Instead of storing the restoration information 43, the standard template storage unit 7
1 stores a standard template. Accordingly, the standard template restoring unit 22 is deleted.

(2) Description of Operation Next, the operation of the second embodiment of the present invention will be described in detail with reference to FIG.

As in the first embodiment, the matching section 13 first performs a large classification matching process. In the large classification matching process, reduced templates are sequentially input from the template storage unit 15 and matching is performed with the reduced pattern created by the input pattern reduction unit 21. When the matching with all the reduced templates is completed, the large classification matching result is output to the large classification result storage unit 24.

Next, the matching section 13 performs a detailed matching process. In the detailed matching process, the large classification result storage unit 2
The template IDs are sequentially input from No. 4, the reduced template corresponding to the template ID is input from the template storage unit 15, and the standard template 41 is input from the standard template storage unit 71, and matching is performed with the input pattern.

According to the second embodiment of the present invention, there is an aspect that the template storage area increases, but the standard template restoring unit 22 becomes unnecessary, in other words, the process of restoring the standard template becomes unnecessary. In other words, there is an effect that the circuit scale can be reduced when the apparatus is configured by hardware or by hardware. Further, in the second embodiment of the present invention, since the reduction ratio is not limited to 1/4 as in the first embodiment, for example, if the reduction ratio is set to 1/16, it is possible to greatly reduce the entire processing time. effective.

[Third Embodiment] Next, a third embodiment of the present invention will be described in detail with reference to the drawings.

(1) Description of Configuration FIG. 9 is a block diagram showing a detailed configuration example centering on the matching unit 13 of the character recognition device according to the third embodiment of the present invention. 9, the matching unit 13 of the character recognition device according to the third embodiment of the present invention includes an input pattern reducing unit 21, a standard template restoring unit 22, a matching processing unit 23, and a standard template buffer 81.

In the third embodiment of the present invention, the large classification result storage section 24 is deleted. Therefore, when the result of the large classification matching process is good, for example, when the number of mismatches is equal to or less than a predetermined value, the detailed matching process is immediately performed on the template. The reduced template buffer 81 is a buffer for temporarily storing the reduced template input from the template storage unit 15 during the large classification matching process. At the time of the detailed matching processing, the reduced template 42 input from the reduced template buffer 81 and the restored information 4 input from the restored information storage unit 16 are used.
3, the standard template restoration unit 22 restores the standard template 41.

(2) Description of Operation Next, the operation of the third embodiment of the present invention will be described with reference to FIGS.
This will be described in detail with reference to FIG.

FIG. 10 is a diagram showing a flow of the operation of the matching unit 13 according to the third embodiment of the present invention. Matching unit 1
3 first performs the large classification matching process 91 on the template I
It starts from D'0 '. If the large classification matching processing result 91 is good, the detailed matching processing 92 is immediately performed.
In the example shown in FIG. 10, since the result of the template ID “A” was good (93), the template ID
The restoration information 43 of “A” is input from the restoration information storage unit 15 (94), and a detailed matching process 92 is performed.

Since the data stored in the reduced template buffer 81 can be used as the reduced template 42 of the template ID 'A', there is no need to input it from the template storage unit 15. When the detailed matching process 92 with one template is completed, the large classification matching process is restarted from the next template. In the example shown in FIG. 10, the template ID is restarted from “A + 1” (9
5). Thereafter, every time a good result is obtained in the large classification matching process, the detailed matching process is repeatedly performed.
Process up to the final template and end.

According to the third embodiment of the present invention, there is an aspect that it cannot be guaranteed that the template for performing the detailed matching process is efficiently limited depending on whether the result of the large classification matching process is good or bad. When the number of templates to be performed can be limited efficiently, there is no need to input a reduced template from the template storage unit 15 at the time of the detailed matching processing, so that the processing can be speeded up.

[0050]

As described above, according to the present invention, means for storing a reduced template obtained by reducing the standard template is provided, and the reduced pattern created from the input pattern to be recognized is matched with the reduced template. Thereafter, control is performed so as to perform matching between the input pattern and the standard template, and at the time of detailed matching processing, control is performed so that information for selecting a processing target template is provided to the matching processing means. Since the standard template is a reduced template while maintaining two-dimensional features, the following effects can be obtained.

The first effect is that the standard template is reduced and the reduced template is stored, so that the template storage area can be reduced. In this case, the entire storage area can be reduced even if restoration information for restoring the standard template is added.

The second effect is that the template for performing the detailed matching processing based on the result of the large classification matching processing which can be executed at high speed is limited, so that the processing time of the matching processing means can be reduced.

The third effect is that since the reduced template retains the two-dimensional characteristics of the standard template, the matching process can be performed in the same manner as the detailed matching process.
A special matching mechanism for the large classification matching process is not required.

[Brief description of the drawings]

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

FIG. 2 is a block diagram illustrating a detailed configuration example centered on a matching unit of the character recognition device according to the first embodiment of the present invention.

FIG. 3 is an explanatory diagram showing a configuration example of a standard template according to the first embodiment of the present invention.

FIG. 4 is an explanatory diagram showing a procedure of reduction and restoration according to the first embodiment of the present invention.

FIG. 5 is an explanatory diagram illustrating an example of creating restoration information according to the first embodiment of this invention.

FIG. 6 is an explanatory diagram showing an operation flow of a matching unit of the character recognition device according to the first embodiment of the present invention.

FIG. 7 is an explanatory diagram showing an example in which a reduced template according to the first embodiment of the present invention has two-dimensional characteristics of a standard template.

FIG. 8 is a block diagram illustrating a detailed configuration example centering on a matching unit of the character recognition device according to the second embodiment of the present invention.

FIG. 9 is a block diagram illustrating a detailed configuration example centering on a matching unit of a character recognition device according to a third embodiment of the present invention.

FIG. 10 is an explanatory diagram showing an operation flow of a matching unit of the character recognition device according to the third embodiment of the present invention.

[Explanation of symbols]

 13 matching unit 15 template storage unit 16 restoration information storage unit 21 input pattern reduction unit 22 standard template restoration unit 23 matching processing unit 24 large classification result storage unit 81 reduced template buffer

Claims (6)

[Claims] 1. A character recognition apparatus for performing character recognition based on template matching, comprising: a reduced template storage unit for storing a reduced template obtained by reducing a standard template; and a reduced template for creating a reduced pattern from the input pattern to be subjected to the character recognition. Means,
A character recognition apparatus comprising: a matching processing unit that performs matching between the reduced pattern created by the reducing unit and the reduced template and then matches the input pattern with the standard template. 2. A restoration information storage unit for storing restoration information for restoring the reduced template to the standard template; a reduced template input from the reduced template storage unit and an input from the restored information storage unit. 2. The character recognition device according to claim 1, further comprising a standard template restoring unit that restores the standard template based on the restoration information. 3. The method further includes temporarily storing a result of a large classification matching process for matching the reduced pattern and the reduced template, and a processing target during a detailed matching process for matching the input pattern and the standard template. 3. The character recognition apparatus according to claim 2, further comprising a large classification result storage unit for giving information for selecting a template to the matching processing unit. 4. A reduced template buffer for temporarily storing the reduced template input from the reduced template storage means during a large classification matching process for matching the reduced pattern with the reduced template, wherein the standard template is restored. Means for restoring the standard template based on the reduced template input from the reduced template buffer and the restoration information input from the restoration information storage means during a detailed matching process for matching the input pattern with the standard template The character recognition device according to claim 2, wherein the character recognition is performed. 5. The character recognition apparatus according to claim 1, wherein the reduced template is a template obtained by reducing the standard template while maintaining two-dimensional characteristics. 6. A character recognition method for performing character recognition based on template matching, wherein a character is cut out and a size of each character is normalized from a target image of the character recognition, and a character recognition method is performed based on the cut and normalized pattern. Creating a reduced pattern, first performing matching between the reduced pattern and a reduced template obtained by reducing the standard template, and then performing matching between the cut-out and normalized pattern and the standard template. Character recognition method.