JPH06259601A - Character recognition device - Google Patents

Abstract

PURPOSE:To raise the probability that characters are correctly recognized by determining the character pattern, where the number of picture elements of line parts is largest, among standard character patterns included in a character picture as a recognized character. CONSTITUTION:The character picture of each character is segmented and is converted to a dot pattern (binary picture) One input character pattern PI as the object of character recognition is selected, and the number Nb of black picture elements of this pattern is calculated, and one standard character pattern PR whose number of black picture elements is smaller than Nb is selected. When AND between the standard character pattern PR and the input character pattern PI is equal to the standard character pattern PR, it is judged that the standard character pattern PR is included in the input, character pattern PI. In this case, the character code of the standard character pattern PR is determined as that of the input character pattern. Thus, the number of standard character patterns used for recognition of character pictures is reduced.

Description

Detailed Description of the Invention

[0001]

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

[0002]

2. Description of the Related Art As a device for recognizing a character from a character image, there are Japanese Patent Laid-Open Nos. 52-42025 and 60-23.
The one described in Japanese Patent No. 0279 is known.
In these devices, the standard character pattern is thickened first
The pattern and the thinned second pattern are prepared, and the character is recognized by comparing the first and second patterns with the character image.

[0003]

In the above-mentioned conventional technique, two character patterns must be prepared for one character, and even if the characters have the same character code but different typefaces, It was necessary to prepare a separate character pattern. For example, one typeface in Japanese contains about 6000 characters, so if a character pattern is prepared for five typeface characters, an enormous number of character patterns of 6000 × 5 × 2 = 60,000 are prepared. There was a need.

In the apparatus disclosed in Japanese Patent Laid-Open No. 52-42025, two character patterns are created by performing a thinning process and a thickening process from the standard pattern, so that the standard pattern should be stored in advance. Although the number of patterns is reduced, there is a problem that thickening processing and thinning processing are required. Further, it is the same as in Japanese Patent Laid-Open No. 60-230279 in that two character patterns are created for one character and compared with a character image.

The present invention has been made to solve the above-mentioned problems in the prior art, and provides a character recognition device capable of reducing the number of standard character patterns used when recognizing a character image. With the goal.

[0006]

In order to solve the above-mentioned problems, an apparatus according to the present invention is an apparatus for recognizing a character from a character image read by an image input apparatus, wherein (A) each of a plurality of characters And (B) comparing the character image read by the image input device with the standard character pattern stored in the storage device, and storing the standard character pattern in the line portion of the character image. Among the characters included in the pattern, a character having the largest number of pixels in the line portion is identified as a recognized character.

[0007]

In the standard character pattern included in the character image, the probability that the character is correctly recognized can be increased by determining the character having the largest number of pixels in the line portion as the recognized character.

[0008]

【Example】

A. Device Configuration: FIG. 1 is a block diagram showing a character recognition system as an embodiment of the present invention. The character recognition system includes a character identification device 10 and an image input device 20. The character identification device 10 is a workstation or a personal computer, and the image input device 2
0 is a scanner.

The character identification device 10 includes a CPU 11 and an RO.
An M12, a RAM 13, a keyboard 14, a display control circuit 15, a CRT 16 and a printer 17 are provided. The character image read by the image input device 20 is stored in the RAM.
13 is stored. A standard character pattern is also stored in the RAM 13. However, the standard character pattern may be stored in another storage means such as a magnetic disk or a magneto-optical disk. By executing the application program stored in the ROM 12, the CPU 11 exerts a function as a character recognition unit that recognizes a character included in a character image.

B. Preparation of Standard Character Pattern: FIG. 2 is an explanatory diagram showing a method of creating a standard character pattern. 2A and 2B are binary images (hereinafter, referred to as "character pattern") P1 of two characters having the same character type but different typefaces.
P2 is shown. Here, the "character type" refers to a character represented by the same character code (JIS code, division mark code, etc.). Two character patterns P in FIGS. 2A and 2B
1 and P2 are defined by a character code and a typeface. In addition, the value "1" of the binary data is assigned to the line portion in the character pattern, and the value "0" is assigned to the background.

When creating a standard character pattern, first, a pattern P of a plurality of characters having the same character type but different typefaces is used.
The logical product of 1 and P2 is taken to obtain a pattern P3 as shown in FIG. The number of character patterns for which the logical product is obtained is not limited to two and may be three or more.

Next, by thinning the pattern P3, the standard character pattern P having a width of one pixel shown in FIG.
Create R. This standard character pattern PR is a standard character pattern common to the two original character patterns P1 and P2.

Depending on the character, the standard character pattern may be created by skeletonizing the character pattern itself without taking the logical product of a plurality of character patterns.

By the way, in the typeface, there is almost a skeleton of a shape, such as the Mincho system (Hosyo-Mincho, Chu-Mincho, Tai-Mincho) or the Gothic system (Thin Gothic, Middle Gothic, Thick Gothic). There are several typeface systems with the same line weight but different line thickness. If standard character patterns are created by the method of FIG. 2 for a plurality of fonts whose shapes are similar to each other, one set of standard character patterns is created for a plurality of fonts. As a result, the amount of standard character patterns can be significantly reduced.

FIG. 3 is a conceptual diagram showing the structure of a standard character pattern file containing the standard character pattern thus created. The standard character pattern file includes a character code, a standard character pattern (binary image), and the number of black pixels in the pattern. As shown in FIG. 3, when a plurality of standard character patterns are required even for the same character "Ka", a plurality of standard character patterns are registered for the same character. Further, in the standard character pattern file, standard character patterns are registered in the order of increasing number of black pixels (the number of pixels having a binary data value of 1). The reason for this will be described later. The standard character pattern file is stored in the RAM 13.

C. Character Recognition Procedure: FIG. 4 is a flowchart showing a procedure for recognizing a character using the character recognition system of FIG. In step S1, the image input device 20
To read the character image printed on the document.

In step S2, the character images are cut out one by one, and the character image of each character is formed into a dot pattern (binary image formation). At this time, a process of removing noise components such as dust and scratches included in the character image is also executed. FIG. 5 (A) shows the input character pattern PI thus obtained.

In step S3, one input character pattern PI which is the object of character recognition is selected, and the number Nb of black pixels thereof is calculated. In step S4, one standard character pattern PR (FIG. 5B) in which the number of black pixels is Nb or less is selected. As shown in FIG. 3, in the standard character pattern file, standard character patterns are arranged in descending order of the number of black pixels.
In step S4, the standard character patterns PR are selected in descending order of the number of black pixels (that is, from the top). For example, when the number Nb of black pixels of the input character pattern is 31, the Gothic type “ka” is first selected, and when step S4 is executed next, the Mincho type “ka” is selected.

In step S5, it is determined whether the line portion (black pixel portion) of the selected standard character pattern PR is included in the line portion of the input character pattern PI. Specifically, the standard character pattern PR and the input character pattern PI
When the logical product of and is equal to the standard character pattern PR, it is determined that the standard character pattern PR is included in the input character pattern PI. Alternatively, it may be determined that the standard character pattern PR is included in the input character pattern PI when the logical sum of the standard character pattern PR and the input character pattern PI is equal to the input character pattern PI.

If it is determined that the standard character pattern PR is included in the input character pattern PI, step S6.
In, the character code of the standard character pattern PR is determined to be the character of the input character pattern.

On the other hand, in step S5, when the standard character pattern PR is not included in the input character pattern PI, the process proceeds to step S7, and when another standard character pattern remains in the standard character pattern file, the step is executed. Returning to S4, the next standard character pattern is selected and step S5 is executed again. When the comparison with all the standard character patterns having the number of black pixels Nb or less is completed in step S7, the character cannot be identified, so that the step S8 is performed.
At, a predetermined unidentifiable code is set for the input character pattern.

When the character code of one input character pattern PI is determined in step S6 or step S8, the process proceeds to step S9. If there is the next input character pattern, the process returns to step S3 and the processes of steps S4 to S8 are performed. repeat. In this way, character codes are determined for all input character patterns read by the image input device 20.

As described above, since the standard character pattern PR included in the input character pattern PI is determined to be the recognized character, the number of black pixels of the recognized standard character pattern is the number of black pixels of the input character pattern PI. It is Nb or less. Therefore, in step S4, the standard character pattern having the number of black pixels equal to or less than the number Nb of black pixels of the input character pattern PI is selected.

Furthermore, since the standard character pattern files of FIG. 3 are arranged in the order of the number of black pixels, the recognized character has the largest number of black pixels among the standard character patterns included in the input character pattern PI. It is a thing. For example, when trying to recognize the input character pattern "5", both of the two standard character patterns "5" and "3" are included in the input character pattern "5". Many characters "5" are recognized characters. Similarly, when trying to recognize the input character pattern "ga", both of the two standard character patterns "ga" and "ka" are included in the input character pattern "ga". The character "ga" with many characters is the recognized character. On the other hand, when trying to recognize the input character pattern "ka", the standard character pattern "ga" is not included in the input character pattern "ka" and the standard character pattern "ka" is included. Is the recognition character. As described above, since the character having the largest number of black pixels among the standard character patterns included in the input character pattern PI is the recognized character, it is possible to correctly recognize the character.

Further, since the standard character pattern is composed of lines having a width of one character, even if the lines of the input character pattern are slightly deviated, the line part of the input character pattern is The line part will be included. Therefore, the probability of correct character recognition is high.

D. Modification: The present invention is not limited to the above-described embodiment, and can be implemented in various modes without departing from the scope of the invention.
For example, the following modifications are possible.

(1) In the above embodiment, when the standard character pattern included in the input character pattern does not exist, the standard character pattern having the highest degree of inclusion of the input character pattern is recognized instead of setting the unrecognizable code. You may make it determine as a character. There are various possible inclusion degrees of the input character pattern. For example, the input character pattern and the standard character pattern may be compared for each row (or each column) of pixels, and the number of rows (or the number of columns) included in the input character pattern may be used as the inclusion degree. At this time, if the number of lines that is not included in the input character pattern exceeds a predetermined number, it is considered that the characters cannot be recognized character candidates, and the processing time can be shortened if the line-by-line comparison is terminated halfway. It is possible. Alternatively, the number of black pixels included in the input character pattern itself may be used as the inclusion degree.

When there is no standard character pattern completely included in the input character pattern, the characters may be displayed in the CRT or the like in descending order of inclusion degree so that the operator can select the correct character.

(2) In some typefaces, horizontal lines are much thinner than vertical lines, such as Mincho typeface. When recognizing characters in such a typeface, only the horizontal lines of the input character pattern may be thickened and then compared with the standard character pattern. The operator can specify whether to thicken the horizontal line.

For example, the input character pattern P of FIG.
Ia is the standard character pattern PR1 shown in FIGS.
Consider the case of comparison with PR2. In this case, FIG. 6 (A)
6A is included in the input character pattern PIa, but is not included in the standard character pattern PR2 of FIG. 6B, the standard character pattern PR of FIG.
The character "1" of 1 is determined as the recognized character. However, FIG.
For the input character pattern PIa of (C), it is preferable that the character “Sat” of the standard character pattern PR2 of FIG. 6B be the recognized character. As shown in FIG. 6C, for an input character pattern PIa with a thin horizontal line, the horizontal line portion of the input character pattern PIa is thickened to the portion shown by the white square in FIG. 6C. In this way, since both of the two standard character patterns PR1 and PR2 are included in the input character pattern PIa, the standard character pattern PR2 having a larger number of black pixels is determined as a recognized character, and correct recognition is performed.

As another example, the input character pattern PIb of FIG. 7C is replaced by the standard character pattern P of FIGS. 7A and 7B.
Consider the case of comparison with R1 and PR2. Also at this time
7C, the input character pattern PIb of FIG.
7C is thickened up to the portion indicated by the white square in FIG. 7C. In this case, the standard character pattern PR1 of FIG. 7 (A) is included in the thickened input character pattern PIb, but the standard character pattern PR2 of FIG. 7 (B) is not included, so the standard character pattern of FIG. 7 (A) is included. The character “±” of the pattern PR1 is determined as the recognized character. Thus, FIG. 7 (A)
It is correctly recognized that the character has the standard character pattern PR1.

As described above, in the case of recognizing characters in a typeface with thin horizontal lines, it is possible to increase the probability of correct recognition by thickening only the horizontal line portion of the input character pattern. As a method of thickening the horizontal line portion of the input character pattern, there are a first method optically performed by the image input device 20 and a second method of correcting image data by the character identification device 10.

In the first thickening method, the light source is scanned in the direction in which the line is thickened (vertical direction in FIG. 7), and the width of the slit provided in the direction perpendicular to the scanning direction of the light source is widened. As a result, blurring occurs in the scanning direction of the light source, and the upper and lower horizontal lines of the character pattern are thickened. The width to be thickened can be adjusted by the width of the slit.

In the second thickening method, the number of black pixels arranged in a direction perpendicular to the direction of thickening the line (that is, in the horizontal direction) is checked, and a predetermined number (eg, 7) or more of black pixels are arranged. Should be recognized as a horizontal line, and black pixels should be added above and below it. At this time, as shown in FIG. 8, when there is only one pixel between the horizontal lines, the pixels sandwiched between the horizontal lines (pixels marked with an X in FIG. 8) remain white pixels. Try to leave. By doing so, it is possible to prevent the space between the horizontal lines from being collapsed.

[0035]

As described above, according to the character recognition device of the present invention, it is possible to reduce the number of standard character patterns used when recognizing a character image.

[Brief description of drawings]

FIG. 1 is a block diagram showing a character recognition system for recognizing characters by applying an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a method of creating a standard character pattern.

FIG. 3 is a conceptual diagram showing the configuration of a standard character pattern file.

FIG. 4 is a flowchart showing a procedure of character recognition.

FIG. 5 is a diagram showing a comparison between an input character pattern PI and a standard character pattern.

FIG. 6 is a diagram showing a first example of an input character pattern and a standard character pattern when a horizontal line is thickened.

FIG. 7 is a diagram showing a second example of an input character pattern and a standard character pattern when a horizontal line is thickened.

FIG. 8 is a diagram showing an example of thickening a horizontal line.

[Explanation of symbols]

 10 ... Character identification device 11 ... CPU 12 ... ROM 13 ... RAM 14 ... Keyboard 15 ... Display control circuit 16 ... CRT 17 ... Printer 20 ... Image input device Nb ... Number of black pixels of input character pattern PI ... Input character pattern PR ... Standard Character pattern

Claims (1)

[Claims] 1. A device for recognizing a character from a character image read by an image input device, comprising: (A) storage means for storing a standard character pattern for each of a plurality of characters;
(B) A character image read by the image input device is compared with a standard character pattern stored in the storage device, and a line portion of the character image includes the line portion of the standard character pattern. A character recognizing device that determines the character having the largest number of pixels as the recognized character.