JPH05290211A - Discrimination method of character kind and the like - Google Patents

Abstract

PURPOSE:To reduce the erroneous discrimination of the character kinds, etc., of the character, etc., of different font even at the time of the character kind and the same character by obtaining the difference between the centroid value and the average value of a frequency by calculating them from an actual size for every character kind, etc., and using a value obtained from a frequency distribution when the difference is smaller than a threshold value and an intermediate value between the centroid/ average values when the difference is larger respectively as standard sizes. CONSTITUTION:A character recognition circuit 21 recognizes N-number of characters. The frequency of height and width obtained concerning the N-number of the characters are read from a size frequency storage part 11. Besides, the total value of height and width and the number of the characters are read from a character kind size storage part 9 to calculate the average value by the character kind. Then, whether the difference between the centroid/average values is within the threshold value or not is judged for every character kind and when it it within the threshold value, the intermediate value between the centroid/average values, for example, is calculated for every character kind to set the calculated value to be the standard size. When the difference of the centroid/average values is other than the threshold value, the centroid value is decided to be the standard size.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of discriminating character types such as uppercase / lowercase letters and kanji / kana characters similar to a character code from a character recognition device.

[0002]

2. Description of the Related Art In recent years, information equipment incorporating a character recognition device has been developed. Japanese Patent Laid-Open No. 2-224084 discloses a method for discriminating between uppercase letters and lowercase letters in a character recognition device.

In the above-mentioned publication, the character code and size of each character of the recognition result are sequentially stored, the character type is discriminated from the character code, and the character has the same pattern width or height as the standard size. Letters, or letters with similar lowercase letters, such as "ya" and "ya," or "engineering" in "industrial" and "e" in "aiue"
Judgment is made by referring to a table that is preset for each character code to determine whether the character has a similar glyph to the kanji and kana (hereinafter referred to as the kanji kana-similar glyph). If it is a glyph character, a mark is added to the stored characters, while the actual size of characters with a standard size is totaled for each character type to obtain the recognition result of one document, and the size is measured for each character type. Determine the standard size for each character type by obtaining the distribution or average value, and set a predetermined threshold to the fixed standard size corresponding to the character type previously marked for the character type, uppercase or lowercase letters or kana kana A method of discriminating a character type for discriminating whether a similar character is a kanji or a kana is disclosed.

[0004]

However, in the above conventional structure, the standard size (standard size for each character type) is used.
It is necessary to decide in advance which characters have the same size, but even if they are the same character, different character fonts have different sizes from the actual standard size. Even if the standard size is calculated by calculating the frequency distribution or the average value, it will deviate from the actual standard size,
It has a problem that it is difficult to discriminate.

The present invention solves the above-mentioned problems of the prior art by reducing erroneous discrimination of character types and the like, such as characters having the same character but different fonts, and significantly improving the discrimination accuracy. The purpose is to provide a determination method.

[0006]

In order to achieve this object, the method of discriminating the character type of the present invention calculates the value of the center of gravity (centroid value) and the average value of the frequency from the actual size for each character type. Then, if the difference is smaller than the threshold value, the value obtained from the frequency distribution is used as the standard size, and if it is larger, the intermediate value between the center of gravity value and the average value is used as the standard size.

Specifically, for a predetermined number of characters, the character code output from the character recognition unit, the center line of the character string, and the height, width, and center point of the character pattern are read, and the uppercase and lowercase letters are read according to the character code. Characters such as Kanji and Kana and similar characters are determined, and for uppercase and lowercase characters of the same shape, the difference between the center line of the character string and the center point of the character pattern is calculated, and the height of the character pattern is calculated. Calculate the center of gravity and the average value of the frequency of the width, compare the value of the center of gravity and the average value, and estimate the standard size for each character type according to this result, and calculate the height or width of the character pattern. It is configured to determine the character type and the like by comparing.

[0008]

With this configuration, the centroid value (centroid value) and the average value of the frequencies are calculated from the actual size for each character type, and the difference between them is calculated. If this difference is smaller than the threshold value, it is calculated from the frequency distribution. The value is the standard size, and if it is larger, the value between the center of gravity value and the average value is used as the standard size, so the error between the calculated standard size and the actual standard size can be reduced, and the discrimination accuracy is significantly improved. Can be improved.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the functional arrangement of a discriminating apparatus for executing a method for discriminating character types and the like according to an embodiment of the present invention. 1 is an image input unit for inputting a document to be read as a binary image by an image scanner or the like, 2 is a character string cutout unit for extracting a character string region from the binary image from the image input unit 1, and 3 is a character string cutout unit The character string center line calculation unit for obtaining the center line of the character string cut out in 2, 4 extracts the characters contained in the character string area, the character pattern is in the character recognition unit 5, and the height and width of the character pattern are the character information. A character cutout unit for storing in the height region and the width region of the storage unit 6, 5 is a character recognition unit for outputting a character code corresponding to the character pattern input from the character cutout unit 4, and 6 is recognition of the character recognition unit 5. A character information storage unit 7 stores the result in the character code area. Reference numeral 7 denotes a character string center line calculation unit 3 and a character cutout unit 4 for calculating the deviation between the center of the character string and the center of the character pattern, and the result is stored in the character information storage unit. 6 A deviation calculation unit to be stored in the deviation value area, 8 is a character code classification unit that determines whether the character code stored in the character information storage unit 6 is a character having both upper case and lower case characters, and 9 is a character. A character type size storage unit that stores the result of the character type size in the code classifying unit 10, an average value calculation unit that calculates an average size for each character type from the data in the character type size storage unit 9, and a character code classification unit 11. A size frequency storage unit for storing the result of the size frequency of the unit 9, 12 is a size frequency storage unit 11
Frequency centroid value calculation unit that calculates the frequency centroid value for each character type from the data of 13, the standard size determination unit 13 that determines the standard size from the results of the average value calculation unit 10 and the frequency centroid value calculation unit 12, and 14 is the standard size An upper / lower case discrimination unit that discriminates whether the recognized character is an upper case letter, a lower case letter or the like based on the standard size from the determination unit 13, and a control unit 15 that controls the above operation.

FIG. 2 is a block diagram showing a hardware configuration of a discriminating apparatus for executing a method of discriminating character types and the like.

Reference numeral 20 is an image scanner for inputting a document image as binary data. Reference numeral 21 is an image scanner 20 for inputting binary data.
A character recognition circuit that outputs a character code after cutting out a character string and characters from the binarized image data, 22 is a RAM including a character information storage unit 6, a character type size storage unit 9, and a size frequency storage unit 11, 23 is a CPU (Central Processing Unit), 24 is a character category size information table 25,
It is a ROM configured by the program storage unit 26.

The character type discriminating apparatus configured as described above will be described with reference to the flowchart of FIG.
FIG. 3 is a flow chart when the method of discriminating the character type and the like of this embodiment is executed.

First, the character recognition circuit 21 recognizes N characters (S1). The character recognition circuit 21 outputs the character code, the center line of the character string, the center point of the character pattern, the height and the width. The outputs from the character recognition circuit 21 for all N characters are stored in the character code area, the character string center line area, the character center point area, the height area, and the width area of the character information storage unit 6 (S2). Next, check the type of each N character (English characters, Hiragana, Katakana, etc.), and for Hiragana and Katakana, the same uppercase and lowercase letters (eg "a" and "a", "ya" and " , Etc.) and check whether the uppercase and lowercase letters are the same as for alphabetic characters (“C” and “c”, etc.), or whether the uppercase and lowercase letters are not isomorphic (“A”, etc.), or Whether or not the uppercase and lowercase letters are not the same shape (such as "a") is checked, and a mark is entered in the corresponding column in the attribute area of the character information storage unit 6 (S3). Next, i is set to 1 (S4), and the attribute, height, and width of the i-th character are read from the attribute area, height area, and width area of the character information storage unit 6, respectively. The value of the height and width of the character type to be increased is incremented by 1, and the height and width values of the i-th character are added to the total value of the height and width of the corresponding character type in the character type size storage unit 9. , 1 is added to the column of the number of characters (S5). The initial value of each column of the size frequency storage unit 11 and the character type size storage unit 9 is 0. The values of the character string center line area and the character center point area of the i-th character in the character information storage unit 6 are read out, the degree to which the center of the character pattern is located below the center line of the character string is calculated, and this value is calculated. The data is stored in the shift value area of the information storage unit 6 (S6). i is N
It is determined whether it is smaller than (S7). If it is smaller, i is incremented by 1 (S8), and the process returns to S5. If they are equal or larger, the process proceeds to S9. The height and width frequencies obtained for N characters are read from the size frequency storage unit 11,
The centroid value is calculated for each character type (S9). The total value of the height and width and the number of characters are read from the character type size storage unit 9, and the average value is calculated for the character type (S10). It is determined for each character type whether the difference between the centroid value and the average value is within a threshold value (S
11). If it is within the threshold value, the process proceeds to S12, and if not, the process proceeds to S13. In S12, for example, an intermediate value between the centroid value and the average value is calculated for each character type, and this is set as the standard size. In S13, the center of gravity value is determined as the standard size.
Next, i is set to 1 (S14). When the attribute area of the i-th character of the character information storage unit 6 is read and the upper and lower case letters are the same shape, the height area or width area of the character information storage unit 6 becomes the standard size determined in S12 or S13. From the relationship between the degree of being larger or smaller than the value obtained by multiplying the value of the character category size information table 25 and the size of the value of the deviation value area of the character information storage unit 6, for example, if the former is considerably large, regardless of the latter. Distinguish it as an uppercase letter, and if the former is large in the middle and the latter is large, it is in lowercase,
When the former is large in the middle and the latter is small, uppercase / lowercase is discriminated using rules such as uppercase letters (S1).
5). It is checked whether i is smaller than N (S16). If equal or larger, i is incremented by 1 in S17 and S1
Returning to 5, the processing is ended when it is smaller.

Next, the method of discriminating character types etc. of the character type discriminating apparatus of the present embodiment will be explained using a concrete example.

FIG. 4 is a storage content layout diagram in which specific example characters are stored in the character information storage unit, and FIG. 5 is a storage content layout diagram in which specific example characters are stored in the size frequency storage unit. 6 is a storage content layout diagram in which characters of a specific example are stored in a character category size information table, and FIG. 7 is a storage content layout diagram in which characters of a specific example are stored in a character type size storage unit. is there.

First, as shown in FIG. 4, the character recognition circuit 21
The character type to which the character code of the character information storage unit 6 storing the output from is checked and the attribute area of the character information storage unit 6 is marked. Next, the difference of the character center point area from the value of the character string center line area of the character information storage unit 6 is written in the error value area.

In order to make the description easier to understand, English characters will be described below. As shown in FIGS. 5 and 6, when looking at the attribute area of the character information storage unit 6, the size frequency storage unit 11 corresponding to the values of the height region and the width region of the character marked in the uppercase column of the alphabet. Increment the values in the height and width columns of the capital letters of (Figure 5). Similar processing is performed for lowercase letters. In addition, as shown in FIG. 7, when the attribute area is viewed in the character information storage unit 6, the values of the height area and the width area of the characters having a mark in the upper case column or the lower case column of the alphabet are respectively classified by the character category size information table 25. The total of the values divided by the value of the corresponding character category and the number of characters are stored in the character type size storage unit 9. From Fig. 5, the height center of gravity of uppercase letters is 50.5 by calculation of (50 * 1 + 51 * 1) / 2, the width is 44.0 in the same manner, and the height is 43.5 and the width is 42.0 for lowercase letters. become. On the other hand, the average value is obtained by dividing the total value shown in FIG. 7 by the number of characters, and for uppercase letters, the height is 51.8 and the width is 42.0.
A small letter gives a height of 51.25 and a width of 52.5. For capital letters, the standard size for height and width is within the threshold,
If the height and width are lower than the threshold for lowercase letters, the standard size for uppercase letters is (50.5 + 51.8) / 2.
= 51.15, (44.0 + 42.0) /2=43.0
Becomes On the other hand, the standard size of English small letters adopts only the center of gravity value, and the height is 43.5 and the width is 42.0. Finally, the values in the height and width areas of the characters marked in both letters in the character information storage unit 6 are compared with, for example, the average value of the standard sizes of uppercase letters and lowercase letters. If it is considerably smaller than the threshold value, if it is larger than the threshold value, it is determined to be a capital letter.

[0019]

As described above, according to the present invention, the value of the center of gravity of the frequency (center of gravity value) and the average value are calculated from the actual size for each character type, and the difference between them is calculated. If it is small, the value obtained from the frequency distribution will be used as the standard size, and if it is large, the intermediate value between the center of gravity and the average value will be used as the standard size, so the error between the calculated standard size and the actual standard size should be small. Therefore, it is possible to realize a method of discriminating character types, etc., which can perform discrimination of character types such as uppercase and lowercase letters and character types of different fonts with high accuracy, with very few misjudgments.

[Brief description of drawings]

FIG. 1 is a block diagram showing a functional configuration of a discriminating apparatus that implements a method for discriminating uppercase / lowercase letters according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a hardware configuration of a discriminating apparatus that implements a method for discriminating character types and the like according to the present embodiment.

FIG. 3 is a flowchart for executing a method for determining a character type and the like according to the present embodiment.

FIG. 4 is a storage content layout diagram in which characters of a specific example are stored in a character information storage unit of the determination device of the present embodiment.

FIG. 5 is a storage content layout diagram in which characters of a specific example are stored in a size frequency storage unit of the determination device of the present embodiment.

FIG. 6 is a storage content layout diagram in which characters of a specific example are stored in a character category size information table of the discrimination apparatus of the present embodiment.

FIG. 7 is a storage content layout diagram in which characters of a specific example are stored in a character type size storage unit of the determination device of the present embodiment.

[Explanation of symbols]

 1 image input unit 2 character string cutout unit 3 character string centerline calculation unit 4 character cutout unit 5 character recognition unit 6 character information storage unit 7 deviation calculation unit 8 character code classification unit 9 character type size storage unit 10 average value calculation unit 11 Size frequency storage unit 12 Frequency centroid value calculation unit 13 Standard size determination unit 14 Upper / lower case discrimination unit 20 Image scanner 21 Character recognition circuit 22 RAM 23 CPU 24 ROM 25 Character category size information table 26 Program storage unit

Claims (1)

[Claims] 1. A character code output from a character recognition unit for a predetermined number of characters, a center line of a character string, a height, a width and a center point of a character pattern are read, and the character code is used to detect uppercase / lowercase letters or kana / kana characters. Judges the character type such as similar glyphs and calculates the difference between the center line of the character string and the center point of the character pattern for characters of the same shape, such as uppercase and lowercase, and the frequency of the height and width of the character pattern. Calculates the center of gravity and the average value, compares the value of the center of gravity and the average value, estimates the standard size for each character type according to this result, and compares this with the height or width of the character pattern A method for discriminating character types, etc., characterized by discriminating the character types.