JPH0458383A - Multi font character recognizing device - Google Patents

Abstract

PURPOSE:To withstand noises and to surely recognize even a graphic with an underline, etc., by performing matching with the picture of a character string for recognition object while scanning a template and repeating the matching processing about the font with high coincidence. CONSTITUTION:The template of the font for the recognition object is stored in a template storage means 4, and template matching is performed by a matching processing means 5 while scanning the template against the picture of the character string for the recognition object. A control means 3 controls the matching processing means 5 to successively perform the matching processing to the font with high coincidence. Thus, the segmenting of the character to be recognized is unnecessitated, and the font is recognized in spite of the condition entering the character to be recognized, and the recognition time is shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a multi-font character recognition device that recognizes which of a plurality of fonts a character on a character string image corresponds to.

[Prior Art] Generally, there are many fonts for the same character. For example, even for kanji, there are fonts such as mesaku, gyosho, and cursive. Furthermore, even if we focus only on the measures taken, various types of fonts are used, such as Mincho font and Gothic font.

Font recognition devices are available to recognize such character types. When recognizing a font with a character string like the one shown in Figure 6, for example, when trying to recognize the character "c" surrounded by a dotted line, it is necessary to use a font with a recognizable character type as shown in Figure 7. Prepare fully described templates for font 1, font 2, font 3, etc., match the template with the target character for all characters of all fonts, and select the font with the highest degree of matching for that character string. It is recognized as a font.

That is, first, the character r CJ to be recognized is extracted, and this extracted character "c" is sequentially matched with the template of the first character a of font 1, the template of the next character S, and so on. . When matching of all characters of font 1 is completed, similar matching is performed for all characters of font 2. Thereafter, matching is performed for all characters of all fonts in the same manner.

Similarly, the remaining characters of the string are rh], 'a'.

Matching is also performed sequentially for 'rJ, 'aJ, and so on.

By performing matching for all fonts of all characters in this way, the font of the character string is recognized.

[Problem to be solved by the invention] However, such conventional methods have the problem that a large amount of templates are matched to all fonts of all characters, resulting in a large amount of data processing and a long processing time. . Furthermore, since it is assumed that the characters to be recognized can be cut out, if noise is mixed in as shown in Figure 8(a), the image No. 43 will be cut out as shown in Figure 8(b).
), and it is not possible to extract the part where noise is mixed, so correct recognition cannot be performed.

Underlined letters as shown in Figure 9, number 10
Similarly, for rotated characters as shown in the figure, correct matching cannot be determined because the characters to be recognized cannot be extracted.

SUMMARY OF THE INVENTION Therefore, the present invention provides a multi-font character recognition device that is capable of recognizing fonts regardless of the conditions in which the characters to be recognized are written, and that also takes a short recognition time.

Means for Solving the Problem] In order to solve the above problem, the multi-font character recognition device of the present invention includes a template storage means for storing templates of a plurality of fonts, and a template of a predetermined character of a predetermined font. and a matching processing means that scans the character string image to be recognized and sequentially determines the matching between the character string image and the template, and selects a font with a high degree of matching according to the output of the matching processing means. , and a control means for controlling the matching processing of the matching processing means.

[Operation] Templates of a plurality of fonts to be recognized are stored in the template storage means, and the matching processing means scans the templates against the image of the character string to be recognized to perform template matching. The control means controls the matching processing means so that matching processing is performed sequentially for fonts with a high degree of matching.

[Embodiment] FIG. 1 is a block diagram showing an embodiment of the present invention. In the diagram, 1 is g! ! 2 is an image memo that stores the output data of the image input section 1; 3 is a CPU (control means) that performs overall control; 4 is the image data of the template; Template memory (template storage means) that stores
, 5 is a matching processing unit (matching processing means) that performs matching between the image data of the character string to be recognized and the template data read from the template memory 4 to calculate the degree of matching of the images at each scanning position.

Note that the CPU 3 stores data in the image memory 2 of the image input section 1.
- Loading data from the image memory 2 into the template memory 4, supplying data from the template memory 4 to the matching processing section 5, etc. - Boat fishing ff1lJ# is performed. In addition, this CPU 3 counts the number of times a certain character matches the recognized character string obtained for several font templates, selects a font for which the number of matches is equal to or greater than a predetermined threshold, and performs matching processing on that font. control so that they are continued sequentially.

6 is an operation panel for performing predetermined operations, and 7 is an external interface unit (external I/F) for exchanging signals with the host unit 8.
It is.

FIG. 2 is a flowchart showing the operation of this device.

In step 100, the image input unit 1 captures an image including a character string to be recognized and images of templates of a plurality of fonts. Next, in step 101, scan flags are turned on for all fonts. Then, in step 102, template data of a predetermined character in a predetermined font is loaded from the template pattern memory 4 into the matching processing section 5.

Thereafter, in step 103, the fully loaded template is scanned on an image map containing the character string to be recognized, and X and Y coordinates having a degree of matching equal to or higher than a threshold value are extracted. In step 104, the extracted pixels are integrated to identify the position and number of target characters, and in step 105, the scan flag of the font is turned off for templates with a small number of characters.

Next, in step 106, it is determined whether the processes from step 102 to step 105 have been completed for all fonts whose flags are on for the currently selected character. If there are any remaining fonts for which these processes have not been completed, the flow returns to step 102, and the above-described processes are performed until the processes from steps 102 to 105 are completed for all flag-on fonts.

In step 106, all flag-on fonts of the selected characters are checked from step 102 to 1.
When it is determined that the processes up to 05 have been completed, the next process is to determine whether or not font recognition has been completed for all characters shown in step 107.

If the font iE ffa for all characters has not been completed, the flow repeats the processing from step 102 to step 107. When font recognition for all characters is completed, the entire process of this flow is ended.

A detailed explanation of the above processing is as follows.

Now, in order to recognize the font type of the character string shown in Figure 3, we will use Figure 4 (a), (b), (c).
, (d).

Assume that there are multiple template fonts as shown in (e). The first character scanned (eg, character a) is selected. Then, one of the plurality of fonts (fonts whose flags are turned on) (for example, font 1) is selected, and template data of the selected character a of that font is successively displayed. Further, as the character string image to be recognized, for example, data in the range shown by the dotted line in FIG. 3 is successively displayed. This range is set according to the size of the template. This range is shown in Figure 5 (as an example)
The result will be as shown in a). In this figure, 10×
The character a is written in a range of 10 pixels. Data in this range and template data are compared.

This comparison operation is performed by determining the degree of matching between corresponding data of 10×10 pixels of the template and data of loXIO pixels in a predetermined range. That is, as shown in FIG. 5(a), the pixel data is composed of a black part (the part marked with * in the figure) and a white part (the part with nothing marked in the figure). When the pixels in the black portion correspond to the template data of logic 1, the two are considered to match. Furthermore, when pixels in the white portion correspond to template data of logic 0, the two are considered to match. When the data in the black part corresponds to logic 0 or the data in the white part corresponds to logic 1, it is determined that the two do not match. Then, the number of matching pixels is counted, and when this count value exceeds a certain threshold, it is determined that the image data in that range matches the template data (it is a character of that template). be done.

When this determination is completed, the range indicated by the dotted line in Figure 3 is 1
It is shifted to the right by a number of pixels, and the data in that range is compared with the template data. Thereafter, in the same manner, the range indicated by the dotted line is sequentially moved horizontally to the right. When the position reaches the right end, the range indicated by the dotted line is returned to the left end, and this time it is moved down by one pixel, and then sequentially moved to the right again. In this way, the entire range of the character string image is scanned. As a result, a character corresponding to the character a of font 1 is found out of the characters in the character string shown in FIG. 3, and the number of characters is counted.

After this process is completed, the next step is to detect how many characters in the character string match the character a of font 2 using the same method. There is no match among the strings in Figure 3, so
In this case, the number of matching characters will be zero. If the number of matching characters does not meet a predetermined standard value (for example, 1),
For font 2, turn off the scan flag. This kind of processing is done in font 3, font 4, font 5, etc.
... until the last font, the scan flag is turned off for fonts that do not match the character a.

When the recognition of the letter a is completed up to the last font,
Next, perform the same process for the characters.

At this time, matching (matching) is omitted for those whose scan flag was turned off when scanning the character a,
Matching is performed only for those whose scan flags are on. In other words, if there are no matching characters in the first comparison, it is recognized that the character string does not belong to that font.

In this way, when the verification of the characters whose scan flags are turned on is completed, the next character C9d+
...and performs matching up to the letter Z. Regarding these, if there is no match, the scan flag for that font is turned off in the same manner as described above.

Then, the font with the greater number of matches at the end of the last character match is determined to be the font of the character string.

Note that although the above explanation is based on a method of matching letters a to Z in order, the recognition time will be shorter if the letters are matched starting from the characters with the highest probability of appearing. Furthermore, to detect the degree of matching, it is also possible to perform differential processing on both the character string image and the template image, obtain the information shown in FIG. 5(b), and count the number of matching pixels corresponding to *. In this case, there is no need to count the number of matches in the white part.

[Effects of the Invention] As explained above, according to the multi-font character recognition device of the present invention, a template is scanned to match a character string to be recognized with an image, and the matching process is repeated for fonts with a high degree of matching. This eliminates the need for cutting out processing, is resistant to noise, and allows reliable recognition even of figures with underlines and the like. Furthermore, since fonts with a high degree of matching are recognized while being narrowed down, there is no need to match all characters, which has the effect of increasing processing speed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the multi-font character recognition device of the present invention, FIG. 2 is a flowchart showing its operation, FIG. 3 is a diagram showing character strings recognized by the present invention, and FIG. Figure 4 is a diagram for explaining the font used in the present invention, Figure 5 is a diagram for explaining the detection of the degree of matching,
Figure 6 is a diagram showing character strings recognized by conventional devices;
The figure is a diagram for explaining the font used in the conventional device, and FIGS. 8 to 10 are diagrams for explaining the cutting operation of the conventional device. 1... Image input section, 2... Image memory, 3... C
PU (control means), 4... template pattern memory (template storage means), 5... matching processing section (matching processing means), 6... operation panel, 7.
...External I/F18...Host part.

Claims (1)

[Scope of Claims] Template storage means for storing templates of a plurality of fonts, and selecting the template of a predetermined character of a predetermined font and scanning it against a character string image to be recognized. It is comprised of a matching processing means that sequentially determines matching with a template, and a control means that selects the font with a high degree of matching in response to the output of the matching processing means and controls the matching processing of the matching processing means. Multi-font character recognition device.