JPH0573716A - English character recognition device - Google Patents

Abstract

(57) [Summary] [Purpose] To improve the recognition rate by accurately extracting words. [Structure] The character recognition unit 5 performs character recognition processing based on the position information of the text area, line area, and character area cut out from the document image input by the image input unit 1, and the result is obtained. Based on the definite word segmentation and the candidate word segmentation obtained by the word segmentation processing unit 7, the segmentation candidate generation unit 8 based on the width of the horizontal gap between the adjacent character regions for the character string of each line. , A plurality of division candidates are generated, and whether all of the obtained candidates are correctly spelled as an English word in the character string obtained by division is compared with the English word spelling dictionary 10 by the English word spelling determination unit 9 judge. [Effect] By using the spelling information of English words in addition to the character spacing information, the word segmentation processing is performed, so that accurate word segmentation can be obtained and the recognition rate is improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an English character recognizing device for recognizing English characters.

[0002]

2. Description of the Related Art In recent years, there has been an increasing demand for using a character recognition device as an input device for a computer or the like, and a character recognition device capable of obtaining a stable recognition result is indispensable for improving the performance of a system such as a computer. There is. In the conventional recognition device, the word segmentation process of an English document is performed only on the basis of image information.

[0003]

As described above, the conventional character recognition apparatus performs word segmentation processing of English documents based only on image information, so that documents with narrow character intervals, italicized documents, etc. When recognizing, the words were erroneously separated, which led to a decrease in recognition rate.

The present invention solves the above-described problem. In addition to character spacing information, word segmentation processing is performed using spelling information of English words, so that documents with narrow character spacing, italicized documents, etc. It is an object of the present invention to provide an English character recognizing device capable of performing highly accurate word segmentation processing.

[0005]

In order to solve the above-mentioned problems, an English character recognition apparatus of the present invention comprises an image input unit for inputting a document to be recognized, and a text area for outputting a text area from the input document image. Graphic part of the image of the character area, and the character part image part which divides the text area into line areas by dividing the sentence area into line areas And a character recognition unit that determines the recognition result based on the degree of similarity between them and the figure features of each character type that were previously obtained as a recognition dictionary are determined as word delimiters based on the interval between adjacent character areas. The defined word break and a candidate word break that becomes a candidate for the word break processing unit, and the character string in the area partitioned by the fixed word break are combined by the candidate word break. A division candidate generation unit that divides into several divisions and generates division candidates, and collates whether all the divided character strings are correct in spelling English words with an English word spelling dictionary that holds the spelling of English words. And a spelling determination unit that uses the divided character string for which correct spelling is obtained as a recognition result.

[0006]

With the above configuration, based on the definite word delimiter and the candidate word delimiter obtained by the word delimiter processing unit, in the division candidate generation unit, the character string surrounded by the definite word delimiter exists between the candidate words. Divide by the combination of delimiters to generate multiple division candidates, and for all the obtained candidates, check whether the character string obtained by division is spelled correctly as an English word by comparing it with an English word spelling dictionary. The spelling determination unit makes the determination, and thus the word segmentation error can be suppressed and the recognition rate can be improved.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing an English character recognition apparatus according to an embodiment of the present invention. In FIG. 1, 1 is an image input unit for inputting a recognition target document as a document image, 2 is a document region cutout unit for finding a set of character strings from the input document image, and outputting a document region, 3 is a horizontal region from the document region. A line area cutout unit that finds a series of characters in a direction and cuts out a series of character strings as a line, 4 is a character area cutout unit that divides each character in the line area and cuts it out as a character region, and 5 is a character region image. The character recognition unit 6 compares the figure features and the figure features of each character type previously obtained as a recognition dictionary and determines the recognition result based on the degree of similarity between them. The character recognition unit 6 holds the previously obtained figure features of each character type. The recognition dictionary 7, which is not a definite word definite as a definite word delimiter based on the interval between adjacent character areas, is not definite, but a candidate word delimiter as a candidate is obtained. A word segmentation processing unit, 8 is a division candidate generation unit that generates a division candidate by dividing the character string in the area partitioned by the fixed word division into several ways according to the combination of candidate word divisions, and 9 is a division candidate Whether the character string of is spelled correctly as an English word is determined by comparing it with an English word spelling dictionary that holds the spelling of the English word, and the spelled word with the correct spelling is used as the recognition result. Judgment part, 10 is an English word spelling dictionary holding spelling information of English words, 11 is an internal bus connecting each part of 1 to 5 and 7 to 9, 12 and 13 are connecting 5 and 6, 9 and 10 respectively It is an internal bus. Regarding the English character recognizing device of the present embodiment configured as described above, the entire flow chart of FIG. 2 and FIG.
The operation will be described below with reference to the conceptual diagram showing the process of the process.

First, a document to be recognized is input as a document image in the image input section 1 (process 14). The input document image is sent to the text area cutout unit 2, and the document area cutout unit 2 is sent.
At, the histogram of black pixels existing in the document image is obtained in the horizontal and vertical directions, the document area is cut out from the distribution, and the position information is stored as internal data (Process 1
Five).

Next, the position information of the text area is sent to the line area cutout section 3 to cut out the line in the text area (process 16). In the line area cutout unit 3, paying attention to the fact that the space between lines is larger than the space between adjacent characters, a series of characters in the horizontal direction is found, and a series of character strings is cut out as a line and cut out in the text area. The position information of all the rows that have been stored is stored as internal data.

Next, the position information of the line area is sent to the character cutting section 4 to perform a character area cutting process for the character area (process 17). In the character cutout unit 4, paying attention to the black pixel connection information in the row area and the change in the vertical histogram, the black pixel connection in the horizontal direction is broken or the histogram value is lower than a certain value. Is divided as a character-to-character separation, a character area is obtained for each character, and the position information is stored as internal data.

Next, the position information of the character area is obtained by the character recognition unit 5.
And recognizes all the characters in the text area (process 18). In the character recognition process, the distribution of black pixels is obtained from the image in the character area as a graphic feature, and the recognition dictionary 6
By comparing the similar graphic features of each character type prepared in advance to the degree of similarity between them,
The character type with the highest similarity is used as the recognition result of the character area.

From the above processing 15 to processing 18,
Based on the position information of the text area / line area / character area and the recognition result of the character area, the word division processing unit 7 performs word division processing for dividing a character string that is a series of lines into words. (Process 19). The word segmentation process is performed paying attention to the horizontal gap between adjacent character regions in the line region, and if the gap is a stable size (constant value 1) or more as the word segment, the gap is determined as the word segment. (Definite word delimiter). If the size cannot be determined, but there is a possibility of word breaks (constant value 2) or more, the gap is set as a word break candidate (candidate word break). The constant value 1 and the constant value 2 are values obtained for each line area by the distribution of the size of the gap between the character areas in the line area.

Based on the fixed word delimiter and the candidate word delimiter, the division candidate generation unit 8 performs a division candidate generation process to divide the character string sequence into words to generate a character string candidate as an English word. Yes (process 20). The division candidate generation process is to obtain a plurality of division candidates by dividing a character string surrounded by definite word divisions by a combination of candidate word divisions existing therebetween.

The process of division is shown in FIG. In FIG.
24 is a line in the input document, 25 is a word segmentation process for 24, and 26 is a character string between fixed word segment 1 and 2 is segmented by a combination of candidate word segmentation to obtain segment candidates. The results are shown. In the case of this example, 3 between If, you and just are provided between the fixed word delimiters 1 and 2.
Although there are two words, the gap between these words is narrower than that between other words, so that it cannot be determined as a word division, and three candidate word divisions are obtained as possible character divisions. In the division candidate generation processing, eight combinations of division candidates are generated by the combination of the candidate word divisions, as shown by 26 division candidate results.

In the English word spelling determination unit, it is determined by checking with the English word spelling dictionary 10 whether or not the character string obtained by division is spelled correctly as an English word for all the candidates obtained above ( Process 21).

In 26 showing the result of the division candidate, the character string of the division candidate 1 is "Ifyoujust", but since such a spelling is not in the English word, the determination result is "error". The character string of the division candidate 2 is “If” and “youju”.
"st" and "If" is correct, but "youjus"
Since "t" is not in the English word, the determination result is "wrong". Similarly, all the division candidates are determined, and all the “If”, “you”, and “just” of the division candidate 4 are spelled correctly, so that they are correct and are output as recognition results.

As described above, the processes 20 to 21 are performed on all the character strings surrounded by the definite word delimiters to obtain the recognition result of all the sentence regions (process 22 and process 23).

[0018]

As described above, according to the present invention, in addition to character spacing information, spelling information of English words is used to perform word segmentation processing, so that a document with narrow character spacing or an italicized document can be used. It is possible to reduce the error in the word segmentation processing and improve the recognition rate.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an English character recognition device according to an embodiment of the present invention.

FIG. 2 is an overall flowchart of character recognition processing in the English character recognition device.

FIG. 3 is a conceptual diagram showing a process of the same process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input section 2 Text area cutout section 3 Line area cutout section 4 Character area cutout section 5 Character recognition section 6 Recognition dictionary 7 Word segmentation processing section 8 Division candidate generation section 9 English word spelling determination section 10 English word spelling dictionary

Claims (1)

[Claims] 1. An image input unit for inputting a document to be recognized,
A text area cutout unit that outputs a text area from the input document image, a line area cutout unit that divides the text area one line at a time and cuts it out as a line area, and a character in the line area one character at a time and cuts out as a character area Adjacent to the character region cutout unit, a character recognition unit that compares the graphic feature of the image of the character region and the graphic feature of each character type obtained in advance as a recognition dictionary, and determines the recognition result based on the degree of similarity between them. Based on the space between character areas, the fixed word breaks that are fixed as word breaks and the candidate word breaks that are candidates for the word break processing unit, and the character strings in the area partitioned by the fixed word breaks are used as candidate word breaks. Divide into several combinations according to the combination, and a division candidate generation unit that generates division candidates, and check whether all the divided character strings are correct as spelling of English words. Ri is determined by matching the English word spelling dictionary that holds, English character recognition device comprising a English word spelling determination unit to recognition result divided character string obtained is correct spelling.