JPH02224084A - Discriminating method for capital letter, small letter and character with shape similar to kanji (chinese character) and kana (japanese syllabary) - Google Patents

Abstract

PURPOSE:To decrease erroneous discrimination by discriminating a capital letter and a small letter by using not only a size of the character but also the center coordinate of the character, when a standard size of the character is almost the same by a kind of the character. CONSTITUTION:Whether a character as a result of recognition is a character having both a capital letter and a small letter or not is decided, and when it is a character having both of them, an external form feature quantity containing its character width, character height and that which is obtained by multiplying them is compared with a threshold for setting the character as a capital letter against its standard character, and a threshold for setting it as a small letter, respectively and decided (determined definitely) as one of a capital letter and a small letter. As for that which is not any of them, that is, a character which cannot be decided by the external form feature quantity, a center line L of the line is derived from the center coordinate of a result of recognition of one line, and by comparing a difference of the center X of the character and the coordinate of the center line L with the threshold determined in advance, a capital letter or a small letter is discriminated. In such a way, an erroneous discrimination can be decreased.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for determining character types (such as uppercase or lowercase) in a character recognition device that recognizes characters such as hiragana and katakana. Incidentally, an example of kana characters having uppercase and lowercase letters is shown in FIG.

[Conventional technology]

Conventionally, for example, when distinguishing between uppercase and lowercase letters, it has been known to provide only a threshold value for determining lowercase letters and to compare this threshold with external features of the characters.

[Problem R that the invention attempts to solve]

However, with this method, the size of katakana is generally smaller than that of kanji, and the ratio also changes depending on the typeface.

For this reason, there is a problem in that a case may occur in which the characters of the same size are uppercase in one typeface and lowercase in another typeface.

Therefore, the problem of this invention is to set separate judgment criteria for uppercase and lowercase letters, and to judge characters in between the judgment criteria based on a threshold value of the deviation of the center coordinates of the character from the center of the line. This aims to improve the discrimination accuracy.

[Means for solving the problem] After normalizing the size of the target character whose standard size is approximately the same regardless of the character type and recognizing the character using the same standard pattern for both uppercase and lowercase letters, In addition to memorizing the center coordinates of the circumscribing frame, it also determines whether the character has both uppercase and lowercase letters, and if it has both, the width of the character.

Multiply the character height, character width, and character height to find the external feature amount including (!-), and use the external feature amount as an uppercase character for a predetermined standard character for each character.

It determines whether it is an uppercase or lowercase letter by comparing it with each threshold value for determining lowercase letters, and if it cannot be determined based on these thresholds, it adds undetermined information to the character, and also confirms one line. Every time the work is completed, the center line of the character line is calculated from the center coordinates of each character in the line including the undetermined character, and the difference between the center coordinates of the undetermined character and the center line is set to a predetermined threshold value. Distinguish by comparing 1 and 7 [Operation] Determine whether the recognized character has both uppercase and lowercase letters, and if it has both, multiply the character width by 9 characters and the height by A threshold value for converting external shape features, including the combined values, into uppercase letters for standard characters,
It is determined (determined) as either an uppercase or lowercase letter by comparing it with the threshold value for lowercase letters, and for characters that are neither of these, that is, characters that cannot be determined based on the external feature amount, the center coordinates of the recognition result of the line -. The center line of the line is determined from the center line, and the coordinate difference between the center of the character and the center line is compared with a predetermined threshold value to determine whether it is an uppercase or lowercase letter. This reduces misclassification and improves discrimination accuracy. .

〔Example〕

FIG. 1 is a flowchart showing an embodiment of the invention, and FIG.
The figure is an explanatory diagram for explaining an example of a horizontally written character group and its center line.

First, the character image data is extracted using known image processing (see ■), and the target characters are recognized using the same known method (
■Reference). Next, save the central locus information of this character,
(See ■).Furthermore, the recognition results show that the target character is an uppercase letter,
Determine whether the character has both lowercase letters (see ■), and if the character has both uppercase and lowercase letters, find the character width, height, and the product of the width and height (see ■) e
Next, compare the target character with a predetermined threshold value for capitalizing one or more standard characters (see ■), and if the result is an uppercase character, it is determined to be an uppercase character.
(See ■), If it is not an uppercase letter, compare it with the threshold value L2 (see ■), and if the result is a lowercase letter, it is determined as a lowercase letter (see ■).On the other hand, if neither can be determined Add information to the effect that it is undetermined (see @).

Repeat the above step ■-[phase] to obtain a recognition result for one line (see ■). Next, if there is an undetermined character in the line (see ■), the approximation formula for the center line of the line, Y=ax+b, is calculated from the center coordinates of each horizontally written character in the line using a known method such as the method of least squares. (see 0), and substitute the X-direction center coordinate Xc of the undefined character into the approximation formula to obtain the Y-direction coordinate Y4. as shown in FIG. 2(o). (see 0). Note that the "x" mark in FIG. 2 indicates the center position of each character, and this coordinate YL and the center coordinate Y in the Y direction of the undetermined character.

Determine the stop difference (Yt, Ye) 2, compare it with a predetermined threshold value for standard characters (see ①), and determine whether it is an uppercase or lowercase letter from the result ([Phase]).

Gui? l), ・In other words, the above difference (~'i, 't
Regarding c), the discrimination is made by utilizing the fact that lowercase letters are larger than uppercase letters.

By the way, in the above, it is assumed that the standard size (standard size) does not change depending on the character type, and uppercase letters,
It is designed to recognize lowercase letters.

However, in many printed documents, the standard size differs depending on the type of character (for example, in a single printed document, kanji are significantly larger than kana). Example of standard size is shown in 7th
As shown in the figure. In addition, there are characters with similar glyph shapes in kanji and kana (hereinafter also referred to as kanji-kana similar glyph-shaped characters). An example thereof is shown in FIG.

Therefore, since such a case cannot be handled using the above method, the following method is used. FIG. 3 is a flowchart for explaining the method in such a case.

First, as in the case of FIG. 1, extract character image data using known image processing (see 5). Similarly, the target character is recognized using a known method (see ■).

Next, the character codes and sizes obtained from the recognition results are sequentially stored in the format shown in FIG. 1. (See ■), the character type of the character is kanji according to its character code.

For distinguishing between katakana, hiragana, 2 alphabetic characters, etc. (see ■), the character has the standard size of the character type, or is not shown in Figure 6 and is 1 or 5.゛Yu2. In the format shown in Figure 5, for example, write in advance whether it is a lowercase character such as ``゛yo'' or a kanji-kana similar glyph character such as ``力'' or ``evening'' (not shown in Figure 8). Please refer to the attribute table T set for each character (see ■2)), and mark the memorized character if it has a lowercase letter or a kanji/kana similar glyph. (See ■), if the character has a standard size (see ■), calculate the size of the character using an appropriate method for each type of character, such as frequency distribution calculation, average value calculation, etc. (see ■) ), - Obtain document recognition results ([
Based on the tally result, calculate the standard size of each character type by, for example, finding the most frequent size from the frequency distribution. Search for the characters with , (see ■)
, Calculate each character type in step ◎1. (confirmed 1.
7) A threshold value is obtained by multiplying the standard size value corresponding to the marked character type by a preset ratio, etc., and compared with the actual character size to determine whether it is an uppercase or lowercase letter. (see 0). Furthermore, the standard size kanji and hiragana obtained in step ■.

Check whether the size is different by comparing the katakana stop difference with the preset + and threshold (see 0), and if it is different, for the kanji-kana similar glyphs with "7-" prefixed. , for all characters similar to this, e.g.
The ratio of the character size in table T shown in the figure to the standard size of the character type to which the character belongs The ratio preset in the table is multiplied by the standard size for each character type obtained in step ■ by 1.3. Estimate the size, compare this with the actual character size, and select the character with the closest size as a candidate (see ■). ”-Repeat step 0-■ until the end of the document (see [phase]). In addition, in determining whether a kanji-kana-like glyph is a kanji or a kana, °ζ determines the character types before and after it. It is desirable to also use a method to do so. In addition, in Sat and Ki, the determination of uppercase or lowercase letters and the determination of whether Kanji-kana-like glyphs are kanji or kana are performed at the same time, but it is also possible to perform only one of them. Of course.

〔Effect of the invention〕

According to this invention (17), when the standard size of characters is approximately the same depending on the character type, uppercase and lowercase letters are determined using not only the size of the character but also the center coordinates of the character (,
Therefore, there is an advantage that misclassification can be reduced and classification accuracy can be improved.

Also, if the standard size of characters differs depending on the character type, calculate (determine) the standard size for each character type.
Therefore, it is possible to improve the accuracy of discrimination between uppercase letters, lowercase letters, and characters with similar glyphs such as kanji, kana, and kanji.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing one embodiment of this invention, FIG. 2 is an explanatory diagram for explaining an example of a horizontally written character group and its center line, and FIG. 3 is a flowchart showing another embodiment of this invention. FIG. 4 is an explanatory diagram for explaining the storage mode of recognition results, and FIG. Figure 6 is an explanatory diagram showing an example of a character attribute table with similar character shapes; Figure 7 is an explanatory diagram illustrating an example of characters with similar shapes in uppercase and lowercase; Figure 7 is an explanatory diagram illustrating an example of the standard size of each character. FIG. 8 is an explanatory diagram for explaining examples of kanji, kana, and similar glyphs. Code explanation ■7... Center line, P... Center position of undefined character,
T...Character attribute table.

Claims (1)

[Claims] 1) After normalizing the size of the target character whose standard size is approximately the same regardless of the character type and recognizing the character using the same standard pattern for both uppercase and lowercase letters, the circumference of each character in the recognition result is determined. It memorizes the center coordinates of the frame, determines whether the character has both uppercase and lowercase letters, and if it has both, calculates the character width, character height, and multiplies the character width and character height. The external shape feature amount including the object is determined, and the external shape feature amount is compared with each threshold value for determining whether the character is uppercase or lowercase for a predetermined standard character for each character to determine whether it is an uppercase or lowercase letter. If it cannot be determined based on these thresholds, the character is given undetermined information, and each time the confirmation process for one line is completed, the character line is calculated from the center coordinates of each character in the line that includes the undetermined character. A method for distinguishing between uppercase and lowercase letters, characterized in that the center line of an undefined character is determined, and the difference between the center coordinates of an undefined character and the coordinates of the center line is compared with a predetermined threshold value. 2) After normalizing the size of target characters, which have different standard sizes depending on the character type, and recognizing characters using the same standard pattern for both uppercase and lowercase characters, the character code and size of each character in the recognition result are sequentially memorized and the character is In addition to determining the character type from the code, it also determines whether the character is a standard size character or a character with a similar shape with lowercase letters by referring to a preset table for each character code. While marking memorized characters, the actual size of standard-sized characters is aggregated for each character type to obtain recognition results for one document, and the frequency distribution or average is calculated from the aggregated value of the measured size for each character type. The standard size is determined for each character type by determining the value, and a predetermined threshold value is set for the determined standard size corresponding to the character type for the previously marked character to determine whether it is an uppercase or lowercase letter. Characteristic method for distinguishing between uppercase and lowercase letters. 3) After normalizing the size of target characters, which have different standard sizes depending on the character type, and recognizing characters using the same standard pattern for both uppercase and lowercase characters, the character code and size of each character in the recognition result are sequentially memorized and the character is In addition to determining the character type from the code, refer to a preset table for each character code to determine whether the character is a standard size character or a character with a similar shape between Kanji and Kana (Kanji and Kana similar shape characters). If it is a kanji, kana, or similar glyph, then the memorized character is marked, while the actual size of standard-sized characters is tallied for each character type to obtain the recognition results for one document, and measured for each character type. A standard size is determined for each character type by determining the frequency distribution or average value from the aggregated size values, and a predetermined threshold value is set for the determined standard size corresponding to the character type for the previously marked character. , a method for determining kanji-kana-like glyph-like characters, which comprises determining whether the kanji-kana-like glyph-like characters are kanji or kana. 4) After normalizing the size of the target characters, which have different standard sizes depending on the character type, and recognizing characters using the same standard pattern for both uppercase and lowercase characters, the character code and size of each character in the recognition result are sequentially memorized and the character is In addition to determining the character type from the code, it also determines whether the character is a standard-sized character, a character with a lowercase letter with a similar shape, or a character with a similar shape between Kanji and Kana (Kanji, Kana, and Similar Characters). By referring to a table set in advance for each code, the memorized characters are marked if they are lowercase characters or characters with similar glyphs to kanji, kana, etc., while the actual size of characters with standard size is determined for each character type. Obtain the recognition results for one document by aggregating them, determine the frequency distribution or average value from the aggregated values of the sizes measured for each character type, determine the standard size for each character type, and then match the previously marked characters to that character type. of uppercase letters, lowercase letters, and Kanji-kana-like glyph-like characters, wherein a predetermined threshold is set on the determined standard size of uppercase letters, lowercase letters, and Kanji-kana-like glyph-like characters to determine whether they are uppercase letters or lowercase letters, or whether the Kanji-kana-like glyph-like characters are Kanji or Kana. Discrimination method. 5) The method for determining uppercase and lowercase letters according to claim 4, wherein in determining whether a kanji-kana-like glyph character is a kanji or a kana, a combination of preceding and succeeding character types is also determined. 6) The method for determining uppercase letters, lowercase letters, and kanji-kana-like glyphs according to claim 5), wherein in determining whether the kanji-kana-like glyphs are kanji or kana, a combination of preceding and succeeding character types is also determined. .