JPH0696277A - Alphabet recognizing device - Google Patents

Abstract

PURPOSE:To provide the alphabet recognizing device which can obtain a result of recognition of high accuracy in a short time by minimizing the influence of a contact character. CONSTITUTION:The device is constituted by providing a character pattern group classifying part 13 for classifying a character pattern into pattern groups by a superposing method, and a contact character pattern group defining part 17 for extracting a word character-string containing the character pattern in the character pattern group decided as a contact character, using other character than a contact character pattern as a key, collating it with a work dictionary 18 and defining a contact character-string.

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 each character constituting an English general document.

[0002]

2. Description of the Related Art In recent years, there is an increasing demand for using a character recognition device as an input device for a computer or the like, and a character input device capable of efficiently obtaining a stable recognition result improves the performance of a system such as a computer. Has become essential.

A conventional English character recognition device will be described below. FIG. 6 is a block diagram showing a functional configuration of a conventional character recognition device.

1 is an image input section for inputting a recognition target document as a binary image, 2 is an image storage section for storing the document image input by the image input section 1, and 3 is one of the document images stored in the image storage section 2. A circumscribing rectangle detection unit that obtains a rectangle circumscribing a character based on a series of black pixels. Reference numeral 4 is a graphic feature extraction that extracts a distribution of black pixels of an image within the rectangle obtained by the circumscribed rectangle detection unit 3 as a graphic feature. The units 5 compare the graphic features obtained by the graphic feature extraction unit 4 with the graphic features of all the recognition target characters prepared in advance, and output the characters having similar features as the recognition result. If there is no such feature, the character recognition unit that determines the image in the rectangle as a contact character, 6 is a recognition dictionary that stores the graphic features of all the characters to be recognized, and 7 is output by the character recognition unit 5. Storing recognition result , 1 8 is focused on the vertical histogram of black pixels in the image determined as a contact character at the character recognition section 5, a partial value is smaller as the contact point between characters
It is a contact character division unit that divides each character.

The operation of the character recognizing device constructed as above will be described below. First, the image input unit 1 inputs a document to be recognized as a binary image, and the image storage unit 2
Remember. Next, the rectangle circumscribing the characters in the document image stored in the image storage unit 2 is determined by the circumscribed rectangle detection unit 3 based on the series of black pixels, and the distribution of the black pixels of the image within the rectangle is graphically displayed. The features are extracted by the graphic feature extraction unit 4. The extracted graphic features are sent to the character recognition unit 5, compared with the graphic features of all recognition target characters prepared in advance as a recognition dictionary, and characters having similar features are output as a recognition result.
If there is no similar feature, the image in the rectangle is determined as the contact character. If the recognition result is obtained by the character recognition unit 5, the result is stored in the recognition result storage unit 7. If not, the image in the rectangle is sent to the contact character division processing unit 8 as the contact character. .

The contact character division processing unit 8 obtains a vertical histogram of black pixels in the pixels, sets a portion having a small value as a contact point between characters, and divides the image at the contact point to detect the contacted characters. Separate one character at a time. The divided images are sent again to the graphic feature extraction unit 4, and after the graphic features are extracted, they are recognized by the character recognition unit 5.

Through the above processing, all characters including contact characters in the document image have been recognized.

[0008]

However, in the above-mentioned conventional configuration, since the recognition result is obtained by performing the recognition process character by character, there is a contact character string, that is, a contact character in the document image. In this case, since the characters are divided one by one based on the figure feature and then recognized, when recognizing a document with many contact characters, low recognition accuracy is obtained due to instability of the division accuracy of the contact characters. It had the problem of not having it. Further, since the recognition processing is performed for each character, a large amount of processing time is required due to a large number of division processing / recognition processing operations, resulting in poor workability.

The present invention solves the above-mentioned conventional problems, and an object of the present invention is to provide an English character recognition apparatus which can minimize the influence of contact characters and can obtain a highly accurate recognition result in a short time. And

[0010]

To achieve this object, the English character recognition apparatus of the present invention summarizes, in summary, a character pattern group classification unit that classifies a character pattern into a pattern group by a superposition method, and determines a contact character. It is configured by a contact character pattern group determination unit that extracts a word character string including a character pattern in the selected character pattern group, and collates it with a word dictionary using the characters other than the contact character pattern as keys to determine the contact character string. . Specifically, an image input unit for inputting a recognition target document, an image storage unit for storing the document image input by the image input unit, and a circumscribed character in the document image stored in the image storage unit A circumscribing rectangle detection unit that obtains a rectangle based on a series of black pixels, a word region cutout unit that obtains a word region based on the horizontal interval of the circumscribed rectangle obtained by the circumscribed rectangle detection unit, and the circumscribed rectangle A character pattern group classification unit that classifies the character patterns in the circumscribed rectangle obtained by the detection unit into a pattern group by the superposition method, and the distribution of black pixels in the circumscribed rectangle obtained by the circumscribed rectangle detection unit as graphic features. The graphic feature extraction unit to be extracted, the recognition dictionary in which the graphic features of all the recognition target characters are stored in advance, and the feature similar to the graphic feature extracted by the graphic feature extraction unit and the recognition dictionary are similar to each other. If a character exists in the recognition dictionary, the corresponding character is stored as a recognition result in the recognition result storage unit described later, and if a similar feature does not exist in the recognition dictionary, it is obtained by the circumscribing rectangle detection unit. A character recognition processing unit that determines an image in a circumscribed rectangle as a contact character in which two adjacent characters are in contact, and a word character string that includes a character pattern in a character pattern group that is determined as a contact character is extracted to make contact. A contact character pattern group confirmation unit that confirms a contact character string by collating with the word dictionary using a character other than a character pattern as a key, and a black pixel of an image of a character pattern group that cannot be confirmed by the contact character pattern group confirmation unit Focusing on the histogram in the vertical direction of, the contact character division processing unit that divides the character having a small value as a contact point between the characters and the character recognition processing unit is obtained. The recognition result storage unit that stores the recognition result and the confirmed character string obtained by the contact character pattern group determination unit, and the result stored in the recognition result storage unit and the word area cutout unit. And a recognition result output unit that outputs the recognition result of the recognition target document based on the word area.

[0011]

With this structure, the representative pattern of the classified character pattern group is recognized, and the contact character is collated with the word dictionary by using the character string having high recognition accuracy in the word string which is not in contact as a key. Since the recognition result is obtained, a highly accurate recognition result can be obtained in a short time.

[0012]

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 a functional configuration of an English character recognizing device according to an embodiment of the present invention, and FIG. 2 is a block diagram showing the device configuration.

In FIG. 1, reference numeral 9 is an image input unit for inputting a recognition target document as a binary image, 10 is an image storage unit for storing the input image, and 11 is a rectangle circumscribing a character in the document image with black pixels. A circumscribed rectangle detection unit that is obtained based on the connection, 1
Reference numeral 2 is a word area cutout unit that obtains a word area based on the obtained horizontal interval of the circumscribed rectangle, and 13 is a method of superimposing the image (character pattern) in the rectangle obtained by the circumscribed rectangle detection unit 11. Character pattern group classification unit for classification, 1
Reference numeral 4 is a graphic feature extraction unit that extracts the distribution of black pixels of the image within the rectangle obtained by the circumscribed rectangle detection unit 11 as a graphic feature, and 15 is the graphic feature obtained by the graphic feature extraction unit 14 and all When a character having similar graphic characteristics is compared as a recognition result by comparing with a recognition dictionary 16 in which the graphic characteristics of the recognition target character are stored in advance, and when there is no character having similar graphic characteristics, the characters in the rectangle A character recognition unit that determines a pattern as a contact character, 16 is a recognition dictionary that stores graphic features of all recognition target characters in advance, and 17 is character recognition in the word area obtained by the word area cutout unit 12. In the unit 15, in the word region including the character pattern in the character pattern group determined to be the contact character, the spelling of the word is stored in advance using the character not determined to be the contact character as a key. A contact character pattern group determination unit that determines a contact character string by referring to 18, a word dictionary that stores the spelling of words in advance, and a character pattern group that cannot be determined by the contact character pattern group determination unit 17. On the other hand, paying attention to the vertical histogram of the black pixels of the image in the rectangle, the portion with a small value is regarded as the contact point between the adjacent characters, and the contact character division processing unit divides the characters one by one. Is a recognition result storage unit for storing the recognition result, and a recognition result output unit 21 for outputting the recognition result of the recognition target document based on the recognition result and the word area cutout result stored in the recognition result storage unit 20.

In FIG. 2, 22 is a scanner for reading a document to be recognized as a binary image, 23 is a central processing unit (hereinafter abbreviated as CPU) for controlling the whole, and 24 is a CPU.
Is a control program for performing overall control, 25 is a recognition dictionary that stores graphic features of all recognition target characters, 2
6 is a word dictionary that stores the spelling of words, 27 is a read-only memory that stores the control program 24, the recognition dictionary 25, and the word dictionary 26, 28 is a document image area read by the scanner 22, and 29 is an image. Circumscribed rectangular area, 30
Is a word area in the image, 31 is a character pattern group, 32 is a recognition result, 33 is a document image area 28, a circumscribed rectangular area 29,
A random access memory for storing the word area 30, the character pattern group 31, and the recognition result 32, and 34 is the CPU 2
A keyboard for externally giving a start / end command to the device 3, 35 is an output device for outputting the recognition result 32, and 36 is an internal bus connecting the scanner 22 to the keyboard 34.

The operation of the English character recognizing device configured as described above will be described below. FIG. 3 is a flowchart showing the overall operation of the English character recognition apparatus of this embodiment.

The binary image of the document to be recognized is input by the image input unit 9 and stored in the image storage unit 10 (step S1),
In the image, a group of black pixels connected in the vicinity of 8 is defined as one character pattern, and a rectangle circumscribing the character pattern is determined by the circumscribing rectangle detection unit 11 and stored as internal data. At this time, if a minute rectangle exists and a rectangle exists in the immediate vicinity of the rectangle in the vertical direction, it is regarded as a separated character such as i, j and integrated (step S2). Also,
A portion of the obtained circumscribed rectangle having a large horizontal interval is obtained as a word segment as a word area and stored as internal data (step S3).

The detected circumscribed rectangle area is sent to the character pattern group classification unit 13, and the character patterns in the circumscribed rectangle are classified into pattern groups by the superposition method and stored in the internal data (step S4).

The circumscribed rectangular area of the representative pattern of the classified character pattern group is sent to the graphic feature extraction unit 14, the image in this area is taken out from the image storage unit, and the distribution of black pixels in the image is used as the graphic feature. Extract (step S5).

The extracted graphic features are sent to the character recognition unit 15 and compared with a recognition dictionary in which the graphic features of all the recognition target characters are stored in advance, and whether the characters have similar features or exist in the recognition dictionary. In this case, the corresponding character is stored in the recognition result storage unit as a recognition result, and when a similar feature does not exist in the recognition dictionary, a contact character in which two adjacent characters touch the image in the circumscribed rectangle (Step S6).

When a contact character is present in the character recognition processing, the contact character processing pattern group determining unit performs the contact character processing (steps S7 and S8).

Next, a method of processing a contact character will be described based on a specific example. FIG. 4 is a flowchart showing a method of processing a contact character, FIG. 5 is a diagram showing a concrete example of a processing target, (a) is a diagram showing an input image, and (b) is a character pattern group classification result. It is a figure which shows (the number under a character is a classification pattern number), and (c) is a figure which shows a character recognition result (* is a contact character).

First, all the words in which the character patterns in the character pattern group determined as the contact character exist are extracted (step S10). A character string other than the contact character is used as a key for each extracted word, and a character string that can be a recognition result of the contact character string is obtained by searching the word dictionary (step S
11). It is determined whether there is a character string that applies to all the words extracted in step S10 (step S1).
2). If there is a character string that applies to all words,
The character string is stored in the recognition result storage unit 20 as the recognition result of the contact character string (step S13). In the example of FIG. 5, [* ey] [* rough] [* air] [ra * e
r] [* an] [* eory] keys are used to perform word matching, and [th] applicable to all words is obtained as a recognition result. If the corresponding character string does not exist in step S12, the representative pattern of the contact character pattern group is sent to the contact character division processing unit 19, and the vertical histogram of the black pixels of the image in the rectangle of this character pattern is displayed. When the obtained value is smaller than the given threshold value, the portion is set as the contact point between the adjacent characters, and the circumscribed rectangular area is divided at the contact point (step S14). Hereinafter, graphic feature extraction processing and character recognition processing similar to those in steps S4 and S5 are performed (steps S15 and S1).
6) The recognition result is stored in the recognition result storage unit 20.

The recognition result stored in the recognition result storage unit 20 and the word region cutout result are output to the recognition result output unit 20.
Then, the recognition result of the recognition target document is output (step S9).

By performing the processing of steps S1 to S9, the character recognition processing of the given document image is performed.

[0026]

As described above, according to the present invention, a character pattern group classification unit for classifying character patterns into pattern groups by a superposition method, and a word character string including a character pattern in a character pattern group determined to be a contact character. By using a character other than the contact character pattern as a key, the contact character pattern group determination unit that determines the contact character string by matching with the word dictionary is provided to recognize the representative pattern of the classified character pattern group and With respect to characters, since a recognition result is obtained by collating with a word dictionary using a character string having high recognition accuracy as a key, it is possible to realize an English character recognition device that can obtain a highly accurate recognition result in a short time.

[Brief description of drawings]

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

FIG. 2 is a block diagram showing a device configuration of an English character recognition device according to an embodiment of the present invention.

FIG. 3 is a flowchart showing the overall control procedure of the English character recognition device in the embodiment of the present invention.

FIG. 4 is a flowchart showing a contact character processing method of the English character recognition device according to the embodiment of the present invention.

FIG. 5 (a) is a diagram showing an example of an input image to be processed by the English character recognition device in one embodiment of the present invention (b) is a diagram showing a classification result of the character pattern group (c) is a character of the character Diagram showing recognition results

FIG. 6 is a block diagram showing a functional configuration of a conventional English character recognition device.

[Explanation of symbols]

 9 image input unit 10 image storage unit 11 circumscribed rectangle detection unit 12 word region cutout unit 13 character pattern group classification unit 14 graphic feature extraction unit 15 character recognition unit 16 recognition dictionary 17 contact character pattern group determination unit 18 word dictionary 19 contact character division Processing unit 20 Recognition result storage unit 21 Recognition result output unit

Claims (1)

[Claims] 1. An image input unit for inputting a document to be recognized, an image storage unit for storing the document image input by the image input unit, and a character circumscribing a character in the document image stored in the image storage unit. A circumscribing rectangle detection unit that obtains a rectangle based on a series of black pixels, a word region cutout unit that obtains a word region based on the horizontal interval of the circumscribed rectangle obtained by the circumscribed rectangle detection unit, and the circumscribed rectangle A character pattern group classification unit that classifies the character patterns in the circumscribed rectangle obtained by the detection unit into a pattern group by the superposition method, and the distribution of black pixels in the circumscribed rectangle obtained by the circumscribed rectangle detection unit as graphic features. The graphic feature extraction unit to be extracted, the recognition dictionary in which the graphic features of all the recognition target characters are stored in advance, and the feature similar to the graphic feature extracted by the graphic feature extraction unit and the recognition dictionary are similar to each other. If a character exists in the recognition dictionary, the corresponding character is stored as a recognition result in the recognition result storage unit described later, and if a similar feature does not exist in the recognition dictionary, it is obtained by the circumscribing rectangle detection unit. A character recognition processing unit that determines an image in a circumscribed rectangle as a contact character in which two adjacent characters are in contact, and a word character string that includes a character pattern in a character pattern group that is determined as a contact character is extracted to make contact. A contact character pattern group confirmation unit that confirms a contact character string by collating with the word dictionary using a character other than a character pattern as a key, and a black pixel of an image of a character pattern group that cannot be confirmed by the contact character pattern group confirmation unit Focusing on the histogram in the vertical direction of, the contact character division processing unit that divides the character having a small value as a contact point between the characters one by one and the character recognition processing unit are obtained. The recognition result storage unit that stores the recognition result and the confirmed character string obtained by the contact character pattern group confirmation unit, and the result stored in the recognition result storage unit and the word area cutout unit. And a recognition result output unit for outputting the recognition result of the recognition target document based on the word area.