JPH03241482A - Dictionary compiling method for character recognizing device - Google Patents

Abstract

PURPOSE:To improve the maintainability of a dictionary for recognition and to dispense with a means for the generation of an original image based on the feature information of the dictionary by preserving an image that becomes the base of feature quantity and the feature quantity of the image in the dictionary for recognition. CONSTITUTION:The dictionary is comprised of three hierarchies of a dictionary file 11, an original image preservation file managing file 12, and an original image preservation file 13. The original image of the feature quantity and the feature quantity of the image in the dictionary are preserved in separate files. Therefore, only the dictionary 11 for recognition can be used by loading on a memory when recognition processing is performed, which enables the same recognition processing as that in the case where no original image is preserved to be performed. Thereby, processing time can be shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for configuring a recognition dictionary in a character recognition device.

[Conventional technology]

In a character recognition device, an input character is identified by using a recognition dictionary in which features of the character are registered in advance for each character code, and determining the distance between the feature amount of the input character and the feature amount of the dictionary. Conventionally, in order to improve the maintainability of this type of recognition dictionary, for example, as described in Japanese Patent Laid-Open No. 63-198180, the feature amounts themselves registered in the dictionary are feature information in a form that can reproduce the original image. When adding or deleting registered characters, there is a method of reproducing the original image based on feature information.

[Problem to be solved by the invention]

In the conventional method described above, the feature values registered in the dictionary are feature information in a form that can reproduce the original image, so the types of feature values that can be used as a dictionary are extremely limited (for example, direction codes cannot be used). In addition, a dictionary whose components are average feature amounts of a plurality of character images cannot faithfully reproduce the original image. Furthermore, means for generating the original image based on the feature information in the dictionary is required.

An object of the present invention is to solve the problems of the conventional method described above and to facilitate maintenance of a recognition dictionary without affecting recognition processing.

[Means to solve the problem]

In order to achieve the above object, claim (1) provides that in a recognition dictionary consisting of records whose components are average characteristic sounds of a single or plural character images, corresponding to each record, The present invention is characterized in that an original image of a characteristic sound and a file for storing the characteristic sound of the image are provided.

In addition, in claim (2), in a recognition dictionary composed of average characteristic sounds of a single character image or a plurality of character images as components, the entire recognition dictionary is divided into parts, and one part includes each characteristic sound used for recognition. are registered in order, and the original image of the characteristic sound and the characteristic sound of the image are stored in order in the other.

[For production]

In order to facilitate the maintenance of the recognition dictionary, it is sufficient to save the images that are the sources of the characteristic sounds in the dictionary. However, in optical character recognition (OCR), this original image data becomes quite large. Accept this data
If a set of characteristic sound and original image is saved for each feature record in a book, a large memory area will be required for recognition processing, or complicated processing will be required to save memory. In that case, there is a risk that the processing time will increase.

For example, in Kanji OCR, if the first and second level characters of JIS are to be recognized, a dictionary of 6353 character types must be provided. If a 64×64 normalized image is stored for each, approximately 3 megabytes of space is required. This original image data is not needed during recognition processing, so if the entire dictionary is read during recognition processing, this area becomes a wasted area. Furthermore, in order to save this area, if a portion of the original image is skipped and loaded onto the memory, 6353 file seeks are required, leading to an increase in processing time.

In claim (1), by saving the original image of the characteristic sound in the dictionary and the characteristic sound of the image in separate files, only the recognition dictionary can be loaded into memory and used during recognition processing. It is possible to perform recognition processing in exactly the same way as when the original image is not saved. In claim (2) as well, it is sufficient to load only the portion of the dictionary in which the characteristic sound used for recognition is registered into the memory, and there is no need for processing such as skipping the 1-source image portion in the recognition process.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows the dictionary structure of the first embodiment of the present invention. In this embodiment, the dictionary structure has three layers: a dictionary file 11, an original image storage file management file 12, and an original image storage file 13. It is.

The dictionary file 11 is a file in which characteristic sounds of character images are registered, and corresponds to a conventional recognition dictionary.

The dictionary file 11 is composed of a plurality of records arranged in the order of record numbers, and one record consists of a header section and a data section. The header contains the record number (&4 in this example),
Character codes and others are stored, and the data section stores the average characteristic sound (for example, a histogram of direction codes) of a single character image or a plurality of character images.

The original image storage file management file 12 is a file that manages the original image storage file 13 that holds character images that are the basis of characteristic sounds to be registered in the dictionary file 11. The records of the management file 12 have a one-to-one correspondence with the records of the dictionary file 11, and like the dictionary file 11, they are arranged in the order of record numbers. One record consists of a header section and a data section, and the header section stores the record number (4 in this example), the number of original images (the number of samples used to take the average, 8 in this example), and other information. The original image file name (oct, 04S in this example) is stored in the data section.

There are as many original image storage files 13 as there are record numbers in the dictionary file 11 and management file 12. One original image storage file 13 consists of a header part and a data part, and the header part contains the file name, the number of original images (8 in this example),
Others are stored in the data section, and data consisting of the original image, its feature amount, and others (creation date, etc.) is stored as many times as the number of original images in the header (8 in this example). The average feature amount of each image in this original image storage file 13 is registered in the dictionary file 11.

These dictionary file 1↓, original image storage file management file 12, and original image storage file ↓3 are stored in an external storage device such as a floppy disk, and during the recognition process, they are used in the conventional case where the original image is not saved. Similarly, only the dictionary file 11 is loaded into the memory and used for recognition processing. Since the recognition process using one dictionary file is the same as the conventional one, further explanation will be omitted.

The dictionary correction process will be explained below.

When correcting the dictionary, the original image is confirmed by referring to the original image storage file 13 using the original image storage file management file 12 corresponding to the dictionary file E1. For example, when checking the original image of the fourth feature (record number 4) in the dictionary file 11, first find the corresponding record in the original image storage file management file 12 using record number 4 as a key, and then The name of the original image storage file stored in the record (in this example, ocr, 04
s) is used as a key to find the original image storage file 13. By extracting and displaying the original image from the original image storage file 3, the original image can be confirmed at a glance.

When deleting features in one dictionary file, the features of the corresponding record in the dictionary file 11 (for example, the record with record number 4) are deleted, and the corresponding record in the original image storage file management file 12 is deleted. The original image storage file name and the original image storage file 13 are deleted. In addition, when deleting a part of the feature values that are the basis of the averaged feature values in the dictionary file 11, the corresponding original image in the original image storage file 13 is displayed and the user confirms the deletion. After that, the original image and its features are deleted, the average is recalculated using the remaining features of the original image, and the feature of the corresponding record in the dictionary file 1 is corrected to the new average value. In this case, it is necessary to change the number of original images in the original image storage file management file 12 and the original image storage file 13.

When adding a new feature to the dictionary file 11, create an original image storage file t3, save the original image and its features for the number of samples, and add the corresponding amount to the original image storage file management file 12. add a record,
Save the original image save file name.

If you want to add new features to a certain feature in the dictionary file and take the average, add the new original image to the corresponding original image storage file 3,
Add the feature amount, calculate the average of all the feature amounts in the original image storage file 3, including the added feature amount, and set the feature amount of the corresponding record in the recognition dictionary file 11 to the newly calculated average value. Fix it. In this case as well, the number of original images in the original image storage file management file 2 and the original image storage file 13 is changed at the same time.

FIG. 2 shows the dictionary structure of a second embodiment of the present invention. In this embodiment, as shown in FIG. 2(a), a dictionary is configured with only a dictionary file 21, and the first half of the dictionary file 21 is used as a recognition dictionary section 22, and the second half is used as an original image storage section 23. All feature amounts used for recognition are registered in the section 22, and the original image and its feature amounts are stored in the original image storage section 23.

As shown in FIG. 2(b), the recognition dictionary section 22 consists of a header section of the entire recognition dictionary group and a plurality of data records. The header section contains the number of features used for recognition (33 in this example).
) and other information are stored. One data record stores a one-character record number code, the number of original images, a feature amount, etc., and there are as many data records as there are feature amounts used for recognition (in this example, there are record numbers 1 to 33). .

The original image storage section 23 includes an offset storage section and an original image/feature amount storage section, as shown in FIG. 2(c). The offset storage unit stores an offset amount up to the original image storage position corresponding to each feature value (data record) of the recognition dictionary unit 22. For example, “65” is the first (
This indicates that the offset amount to the original image corresponding to the feature amount of record number 1) is 65 bytes. Similarly, r4134J indicates that the offset amount to the original image corresponding to the 33rd (record number 33) feature amount is 4134 bytes. The original image/feature amount storage section stores a plurality of sets of original images and their feature amounts corresponding to each feature amount of the recognition dictionary section 22.

In the recognition process, the record length of the recognition dictionary unit 22 is calculated from the number of recognition minute quantities (33 in this example) held in the dictionary header of the recognition dictionary unit 22, and that many data records are stored in the memory. Load into. Therefore, processing such as skipping the original image portion is not necessary, and the same processing as conventional recognition processing can be performed.

The dictionary correction process is performed as follows. For example, to refer to the original image corresponding to the first feature quantity in the recognition dictionary unit 22, the corresponding original image is selected from the offset (65 in this example) stored first in the offset storage unit in the original image storage unit 23. Extract and display images. Since each data code in the recognition dictionary 22 stores the number of original images (number of samples) corresponding to the relevant trace amount, it is only necessary to extract the original images for this number from the offset byte position.

When deleting a feature amount in the recognition dictionary, delete the corresponding trace amount in the recognition dictionary section 22, delete the original image and feature amount in the corresponding part of the one-source image storage section 23, and at the same time delete the feature amount from the offset memory. Correct the entire offset value of the section.

In addition, when deleting a part of the feature quantity that is the basis of the averaged feature quantity, the corresponding original image is displayed, and after obtaining confirmation from the user, the corresponding original image in the original image storage section 23 is deleted. The original image and its features are deleted, the average is recalculated using the remaining features, and the corresponding feature in the recognition dictionary section 22 is corrected to the new average value. In this case as well, the offset value is corrected.

When adding features to the recognition dictionary, add the features (add data records) to the recognition dictionary section 22, and at the same time add the original image and its features to the original image storage section 23, When correcting the offset value, or adding new feature values to a certain feature value in the recognition dictionary unit 22 and taking the average thereof, the corresponding part of the one-source image storage unit 23 After adding the newly added original image and its features, and correcting the offset, calculate the average of all target features in the original image storage unit 23, including the features of the added original image. , the corresponding feature amount of the recognition dictionary unit 22 is corrected to the newly determined average value. [Effects of the Invention] According to the dictionary construction method of the present invention, the following effects can be obtained.

(1) In claims (1) and (2), by saving the image that is the source of the feature amount in the recognition dictionary and the feature amount of the image, the maintainability of the recognition dictionary is improved, and the dictionary There is no need for a means to generate an original image based on feature information, and there are no restrictions on the use of feature amounts.

(2) In claim (1), by saving the original image of the feature amount in the dictionary and the feature amount of the image in separate files,
During the recognition process, only the recognition dictionary can be loaded into the memory and used, and the recognition process can be performed in exactly the same way as when the original image is not saved.

(3) In claim (2), each feature used for recognition, its original image, and the feature of the image are stored separately on the recognition dictionary, thereby allowing processing such as skipping the original image portion. This also eliminates the need for a file to manage the original image storage file.

[Brief explanation of drawings]

FIGS. 1 and 2 are diagrams showing one embodiment of a dictionary structure according to the present invention. 11...Dictionary file, 12...Original image storage file management file, 13...
・Original image storage file, 21... Dictionary file, 22... Recognition dictionary section, 2
3...Original image storage section. Figure 1

Claims (2)

[Claims] (1) In a recognition dictionary consisting of records whose components are average feature amounts of a single or multiple character images, corresponding to each record, the original image of the feature amount and the feature amount of the image 1. A method for configuring a dictionary for a character recognition device, characterized in that a file for storing a file is provided. (2) In a recognition dictionary composed of the average feature amount of a single or multiple character images as a component, the entire recognition dictionary is divided into two, and each feature amount used for recognition is registered in order in one half, On the other hand, a dictionary configuration method for a character recognition device, characterized in that the original image of the feature amount and the feature amount of the image are stored in order.