JP2985243B2 - Character recognition method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method for recognizing a sentence expressed by arranging a plurality of words composed of a plurality of characters such as English characters.

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 such as a computer, and a character recognition device capable of efficiently obtaining a stable recognition result has become indispensable for improving the performance of a system such as a computer. ing. As a conventional character recognition method, a text area and a character area are sequentially cut out from a document to be recognized, and a graphic feature is extracted from image data of each character existing in the character area, and the recognition is performed by comparing with a prepared dictionary. Some have obtained results.

[Problems to be Solved by the Invention] In the conventional character recognition method described above, recognition processing is performed by focusing on each character, and the accuracy of the processing of cutting out a character region is one of the major factors that determine recognition performance. It is a factor. However, in a general document, there is a lot of contact between characters due to noise or crushing, so that it is difficult to cut out a character region accurately, leading to a decrease in recognition rate.

Means for Solving the Problems In order to solve the above problems, the present invention reads a character image to be recognized, cuts out a word from the read character image data, and sets the features and words of the image data of each cut out word. Performs word-by-word recognition by comparing with a dictionary, and for words that cannot be recognized by the word dictionary, cuts out characters constituting the word, and compares the features of the image data of each cut-out character with the character dictionary. Then, the recognition is performed in units of characters.

[Operation] In the present invention, when character recognition processing in units of words and character recognition processing in units of characters are used in combination, efficient character recognition processing can be performed, so that processing speed can be improved. .

Hereinafter, one embodiment of the present invention will be described with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of a character recognition device using the character recognition method of this embodiment. In FIG.
An image input unit 1 inputs a recognition target document as a document image. Reference numeral 2 denotes a text area cutout unit that finds a group of character strings from the input document image and outputs a text area. Reference numeral 3 denotes a word cutout unit that finds a word-unit delimiter from a document area and outputs a range of one word as a word area. Reference numeral 4 denotes a character cutout unit that finds a character unit delimiter from a word area and outputs a range of one character as a character area. Reference numeral 5 compares a graphic feature extracted from the word area with a limited word feature dictionary 7 created based on the graphic feature of a word limited by the frequency of appearance, and outputs a word whose feature matches as a recognition result. It is a recognition processing unit. 6 compares a graphic feature extracted from the character area with a character feature dictionary 8 created based on the graphic features of all characters to be recognized.
This is a secondary recognition processing unit that uses a character with a matching feature as a recognition result. Reference numeral 7 denotes a limited word feature dictionary created based on the graphic features of words limited by appearance frequency. Reference numeral 8 denotes a character feature dictionary created based on the graphic features of all the characters to be recognized. Reference numeral 9 denotes an internal bus that connects each unit such as the image input unit 1 and the text cutout unit 2 as shown in the figure. Reference numeral 10 denotes an internal bus that connects the primary recognition processing unit 5 and the limited word dictionary 7. 11 is 2
An internal bus that connects the next recognition processing unit 6 and the character feature dictionary 8.

FIG. 2 shows a flow chart of the entire processing and FIG. 3 shows a specific example of the character recognition device using the character recognition device of the present embodiment configured as described above, and the operation will be described below. . A document to be recognized is input as a document image by the image input unit (process 12). The input document image is sent to the text region cutout unit 2, and the text region cutout unit 2 obtains histograms of the vertical and horizontal black pixels of the document image,
The sentence area is found based on this. Then, the position information of the text area is stored as internal data (process 13). The position information of the sentence area is sent to the word cutout unit 3, and the word cutout processing is performed on the inside of the sentence area. The word extraction unit 3 focuses on the fact that the space before and after the word is larger than the space between the characters in the word, cuts out a character string sandwiched between spaces having a certain width or more as a word, and obtains position information of all the words in the text area. Is obtained and stored as internal data (process 14). When 24 in FIG. 3 is taken as the result of extracting the sentence area, the result of word extraction is shown in 25 in FIG. Hereinafter, the processing 15 to the processing 22 will be described on the assumption that the position information of the obtained word is repeatedly performed for all the words from the upper left corner of the text area. The position information of the word obtained by the word extraction processing is sent to the primary recognition processing unit 5, and the primary recognition processing unit 5 obtains a graphic feature in the word region determined by the position information, and limits the figure feature in the descending order of appearance frequency. The primary words are compared with a limited word feature dictionary 7 created based on the graphic features of the selected words, and words having matching features are used as the recognition results in the primary recognition processing unit 5. If there is no matching word, it is rejected, that is, rejected and notified to the subsequent stage (process 15). The limited word dictionary is created by limiting the order of appearance frequency. Here, special verbs, auxiliary verbs, personal pronouns, articles, and prepositions of up to three characters, which are generally considered to be frequently appearing, are created as limited words. In FIG. 3, reference numeral 26 denotes the result of the primary recognition processing, and the values of 'she', 'is',' t
The words "he", "of", and "my" were recognized because they were included in the limited words, and were rejected because they did not include "daughter" and "friend". If the primary recognition processing unit 5 recognizes the image, the recognition result is output, and the process proceeds to the process 21 (processes 16 and 22). If rejected, the following processing is performed (processing 16). The word rejected by the primary recognition processing unit 5 is sent to the character cutout unit 4, and is separated one by one by the change in the histogram of black pixels in the vertical direction, and the position information of all the characters existing in the word area is obtained (processing 1
7). In FIG. 3, reference numeral 27 denotes a result obtained by cutting out characters. The obtained character position information is sent from the left end to the secondary recognition processing unit 6, where a recognition process is performed for each character. The secondary recognition processing unit 6 extracts the graphic information of the character based on the sent positional information of the character, and a character characteristic dictionary 8 created based on the graphic information of all the characters to be recognized. Are compared, and a character having a matching feature is output as a recognition result (processes 18, 19). Processing 18 to processing 20 are repeated the number of times of the total number of characters obtained in processing 17 to complete recognition of one word. The result of the secondary recognition processing is shown at 28 in FIG. that's all,
By repeating the processing 15 to the processing 22 for all the words from the upper left corner of the text area, all the words in the text area are recognized and the recognition result in the primary recognition processing and the secondary By combining the recognition results in the recognition processing, a final recognition result can be obtained. The recognition result is shown in FIG. In the figure, 'she', 'is', 'the', 'of',
'my' is the primary recognition processing result, 'daughter', 'frie
nd 'is the result of the secondary recognition processing.

[Effects of the Invention] As described above, the present invention reads a character image to be recognized, cuts out a word from the read character image data, and compares the feature of the image data of each cut out word with the word dictionary. For words that could not be recognized by the word dictionary, the words constituting the words were cut out, and the characters of the cut-out characters were compared with the characteristics of the image data of each character and the character dictionary. In this configuration, efficient character recognition processing can be performed when character recognition processing on a word basis and character recognition processing on a character basis are used together, thereby improving processing speed and achieving stable recognition. The result can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a character recognition apparatus using a character recognition method according to an embodiment of the present invention, FIG. 2 is a flowchart showing a control procedure of this embodiment, and FIG. FIG. 9 is an explanatory diagram showing a target character string and a recognition result. 1. Image input unit 2. Text region cutout unit 3.
Word extraction unit, 4 ... character extraction unit, 5 ... primary recognition processing unit, 6 ... secondary recognition processing unit, 7 ... limited word feature dictionary, 8 ... character feature dictionary,

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G06K 9/62 G06K 9/72

Claims (1)

(57) [Claims] 1. A character image to be recognized is read, a word is cut out from the read character image data, and a feature of the cut out image data of each word is compared with a word dictionary to perform word-by-word recognition. For a word that could not be recognized by the word dictionary, characters constituting the word are cut out, and the feature of image data of each cut out character is compared with the character dictionary to perform character-by-character recognition. Recognition method.