JPH06119484A - Character recognizing device - Google Patents

Abstract

PURPOSE:To provide the high-workability character recognizing device for simultaneously correcting the recognized result of the same character in characters at low recognizing certainly. CONSTITUTION:This device is provided with an image input part 1 for converting a recognizing object document to image data, circumscribed rectangle detection part 3 for detecting any rectangle circumscribed to characters in the image data, a character pattern group classification part 10 for classifying character patterns in the circumscribed rectangle into character pattern group corresponding to an overlap method, graphic feature extraction part 12 for extracting the distribution of black picture elements in the representative pattern of the character pattern group, recognition dictionary 5 storing the graphic features of the recognizing object character, character recognition part 6 for calculating the recognized result by comparing the graphic feature with the recognition dictionary 5, recognized result storage part 7 for storing the recognized result, simultaneous recognized result correction part 13 for correcting the recognized result of the same character pattern group of the low recognition certainly character by inputting the proper recognized result to that character, and recognized result output part 9 for outputting the corrected recognized result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is used for printing characters, dot characters and handwritten character patterns of newspapers, magazines, novels, etc. according to JIS.
The present invention relates to a character recognition device for converting code information such as a code.

[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 recognition device capable of efficiently obtaining a stable recognition result improves the performance of a system such as a computer. Has become essential.

A conventional character recognition device will be described below. FIG. 7 is a functional block diagram of a conventional character recognition device.

Reference numeral 1 is an image input unit for inputting a recognition target document as a binary image and converting it into image data, and 2 is an image data storage unit for storing the image data output from the image input unit 1.
Reference numeral 3 denotes a circumscribing rectangle detection unit that detects a rectangle circumscribing a character in the image data stored in the image data storage unit 2 based on a series of black pixels. Reference numeral 4 indicates a circumscribed rectangle detected by the circumscribed rectangle detection unit 3. Feature extraction unit for extracting the distribution of black pixels as the feature features, 5 is a recognition dictionary that stores the feature features of all recognition target characters in advance, and 6 is the feature feature and recognition dictionary extracted by the feature feature extraction unit 4. 5 is a character recognition unit that compares characters with 5 as a recognition result, and 7 is a character recognition unit 6.
A recognition result storage unit for storing the recognition result obtained in step 8; a recognition result correction unit 8 for inputting a correct recognition result for a character with low recognition accuracy stored in the recognition result storage unit 7 and correcting the recognition result; A recognition result output unit 9 outputs the corrected recognition result stored in the recognition result storage unit 7.

The operation of the character recognizing device constructed as above will be described below. First, the image input section 1
The document to be recognized is input as a binary image, converted into image data, and stored in the image data storage unit 2.

Next, the circumscribing rectangle detection unit 3 detects a rectangle circumscribing a character in the image data stored in the image data storage unit 2 based on a series of black pixels, and the figure feature extraction unit 4 circumscribes the rectangle. The distribution of black pixels in the rectangle is extracted as a graphic feature.

Next, the character recognition unit 6 compares the graphic features extracted by the graphic feature extraction unit 4 with the recognition dictionary 5 and sets the characters having similar features as a recognition result, and the recognition result storage unit 7
To store.

Next, the recognition result correction unit 8 displays the recognition result stored in the recognition result storage unit 7 and asks the user to confirm the recognition result for a character having a low recognition accuracy. If it is, the user corrects the recognition result by inputting the correct character.

Next, the recognition result output unit 9 outputs the corrected recognition result stored in the recognition result storage unit 7.

[0010]

However, in the above-mentioned conventional configuration, since the recognition result is individually corrected for all the characters having low recognition accuracy, the same character exists among the characters having low recognition accuracy. Even in such a case, the recognition result must be corrected character by character, which is complicated and lacks workability.

The present invention solves the above-mentioned conventional problems, and when there are the same characters among the characters with low recognition accuracy, it is possible to collectively correct the recognition results, which is excellent in workability. An object is to provide a character recognition device.

[0012]

In order to achieve this object, a character recognition apparatus of the present invention comprises an image input section for inputting a document to be recognized and converting it into image data, and an image data output from the image input section. An image data storage unit to store, a circumscribed rectangle detection unit that detects a rectangle circumscribing a character in the image data stored in the image data storage unit based on a series of black pixels, and a circumscribed rectangle detection unit. A character pattern group classification unit that classifies the character patterns in the circumscribed rectangle into a character pattern group by the superposition method, a character pattern group classification result storage unit that stores the character pattern group classified by the character pattern group classification unit, and a character pattern. A graphic feature extraction unit that extracts the distribution of black pixels in the representative pattern of the character pattern group classified by the group classification unit as a graphic feature, and a graphic feature of all recognition target characters in advance. A recognition dictionary that stores the recognition result, a character recognition unit that obtains a recognition result by comparing the figure feature extracted by the figure feature extraction unit with the recognition dictionary, and a recognition result storage unit that stores the recognition result obtained by the character recognition unit. , Input a correct recognition result for a character with low recognition accuracy stored in the recognition result storage unit and display all recognition results of the same character pattern group as that character stored in the character pattern group classification result storage unit. It has a configuration including a recognition result batch correction unit for correction and a recognition result output unit for outputting the corrected recognition result stored in the recognition result storage unit.

[0013]

With this configuration, the characters in the image data are classified into character pattern groups, and after the character recognition is performed, a correct recognition result is input for a character having a low recognition accuracy and the same character pattern group as that character is recognized. By correcting all the recognition results, it is possible to collectively correct the recognition results of the same character among the characters with low recognition accuracy.

[0014]

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

FIG. 1 is a functional block diagram of a character recognition device according to an embodiment of the present invention, and FIG. 2 is a device block diagram of the character recognition device.

1 is an image input unit, 2 is an image data storage unit,
3 is a circumscribed rectangle detection unit, 5 is a recognition dictionary, 6 is a character recognition unit,
Reference numeral 7 is a recognition result storage unit, and 9 is a recognition result output unit. Since these are similar to those in the conventional example, the same reference numerals are given and description thereof is omitted.

Reference numeral 10 is a character pattern group classification unit for classifying the character patterns in the circumscribed rectangle detected by the circumscribed rectangle detection unit 3 into a character pattern group by the superposition method, and 11 is a character classified by the character pattern group classification unit 10. A character pattern group classification result storage unit for storing the pattern group, 12 is a graphic feature extraction unit for extracting the distribution of black pixels in the representative pattern of the character pattern group classified by the character pattern group classification unit 10 as a graphic feature, and 13 is The recognition result stored in the recognition result storage unit 7 is displayed and the user's confirmation is requested for a character with low recognition accuracy. In case of misrecognition, the user inputs the correct recognition result and the character pattern group classification result It is a recognition result batch correction unit that corrects all recognition results of the same character pattern group stored in the storage unit 11 as the character.

In FIG. 2, reference numeral 14 is a scanner for reading a document to be recognized as a binary image and converting it into image data.
Reference numeral 5 is a central processing unit (hereinafter abbreviated as CPU) that performs all processing, and 16 is a start of recognition of the CPU 15 from outside
A keyboard 17 for inputting an end instruction and a recognition result to be corrected, a control program area 18 for storing a program describing an instruction given to the CPU 15, and a recognition dictionary area 19 for storing the contents of the recognition dictionary 5. A read-only memory 20 includes an image data area 21 for storing image data output from the scanner 14, a circumscribing rectangular area 22 for storing the circumscribing rectangle detected by the circumscribing rectangle detecting unit 3, and a character pattern group classification. Character pattern group area 2 in which the character pattern group classified by the unit 10 is stored
3, a random access memory including a recognition result area 24 in which the recognition result is stored, 25 is a display device for displaying the recognition result, and 26 is an internal bus connecting the CPU 15 and the scanner 14 or the like.

The operation of the character recognition apparatus of this embodiment having the above-described structure will be described below.

FIG. 3 is a flow chart of the character recognition device in this embodiment. First, a document to be recognized is input by the image input unit 1, converted into image data, and stored in the image data storage unit 2 (S1).

Next, the circumscribed rectangle detection unit 3 detects a rectangle circumscribing the character pattern by using a set of black pixels connected in the vicinity in the image data as one character pattern and stores it as internal data ( S2).

If a small rectangle exists and a rectangle exists in the vicinity of the rectangle in the vertical direction, i. It is regarded as a separated character such as j and integrated.

Next, the character pattern group classification unit 10 classifies the character patterns in the circumscribed rectangle detected by the circumscribed rectangle detection unit 3 into character pattern groups by the superposition method and stores them in the character pattern group classification result storage unit 11. Yes (S3).

Next, the distribution of black pixels in the representative pattern of the character pattern group classified by the character pattern group classification unit 10 is extracted from the image data storage unit 2 by the graphic feature extraction unit 12 as a graphic feature (S4).

Next, the character recognition unit 6 compares the graphic features extracted by the graphic feature extraction unit 12 with the recognition dictionary 5 and stores characters having similar features in the recognition result storage unit 7 as a recognition result ( S5).

Next, the recognition result batch correction unit 13 displays the recognition results stored in the recognition result storage unit 7 on the display device 25.
Is displayed to confirm the recognition result to the user for characters with low recognition accuracy. When the character is erroneously recognized, the user inputs the correct recognition result from the keyboard 16, all character patterns belonging to the same character pattern group as the character are extracted from the character pattern group classification result storage unit 11, and the extracted character is extracted. The pattern recognition result is corrected to the input recognition result (S6).

Next, the recognition result output unit 9 outputs the corrected recognition result stored in the recognition result storage unit 7,
Obtain the final recognition result for the document to be recognized (S
7).

Here, a specific example of the operation of the character recognition apparatus of this embodiment will be described using the recognition target document shown in FIG.

FIG. 4A shows a document to be recognized. First, a recognition target document is input, circumscribing rectangle detection processing is performed, and character pattern group classification processing is performed (S1 to 3).

FIG. 4B shows the character pattern group classification processing result for the recognition target document shown in FIG. 4A.

FIG. 4B is a diagram showing the result of character pattern group classification processing. Here, the upper part of each column is a character pattern,
The lower row shows character pattern numbers. The character pattern numbers are assigned in order from the beginning of the sentence, the character pattern number 1 is assigned to the pattern group of the character pattern "T", and the character pattern number 22 is assigned to the pattern group of the character pattern "th". The recognition target document shown in FIG. 4A is classified into 25 character pattern groups.

Next, graphic feature extraction processing is performed on the representative pattern of the classified character pattern group, and character recognition processing is performed (S4-5).

FIG. 5 shows the character recognition result for the character pattern group classification processing result shown in FIG. 4 (b).

FIG. 5 is a diagram showing the result of character recognition. Here, the upper row of each column shows the recognition result, and the lower row shows the character pattern number. The recognition result “*” is a symbol indicating that there is no similar character in the recognition dictionary 5. The shaded characters indicate that the recognition accuracy is low.

Next, recognition result batch correction processing is performed (S
6). First, the character recognition result shown in FIG. 5 is displayed, and the correction of the user is requested for the recognition result “l” of the character pattern “l” of the character pattern number 14 having low recognition accuracy. When the user inputs from the keyboard 16 that the correct recognition result for the character pattern number 14 is "1", the recognition results of all the character pattern groups classified into the character pattern number 14 are corrected to "1".

FIG. 6A shows a state in which the recognition result batch conversion process is being performed at this time for the character recognition result shown in FIG.

FIG. 6A is a diagram showing a state in which the recognition result batch conversion process is being performed. Here, “platitudinous” and “by-a” in the recognition result are input by one time.
nd-large "," only "," success "
"ful", "real", "problems", "d"
The recognition results of the seven character patterns “l” included in “evolution” are collectively corrected to “l”.

Next, character pattern number 2 with low recognition accuracy
The user is requested to correct the recognition result “*” of the character pattern “th” of No. 2. When the user inputs from the keyboard 16 that the correct recognition result for the character pattern number 22 is "th", the recognition results of all the character pattern groups classified into the character pattern number 22 are corrected to "th".

Here, the recognition result of the six character patterns “th” in the recognition result is collectively corrected to “th” by one input.

Next, recognition result output processing is performed (S7).
The corrected recognition result output to FIG.6 (b) is shown.

FIG. 6B is a diagram showing the corrected recognition result. As is clear from FIG. 6B, it can be understood that the recognition target document shown in FIG. 4A is correctly recognized by the above processing.

[0042]

As described above, according to the present invention, the characters in the image data are classified into the character pattern group, and after the character recognition is performed, the correct recognition result is input to the character with low recognition accuracy and the character is recognized. By correcting all the recognition results of the same character pattern group as the input one, it is possible to collectively correct the recognition results of the same character among the characters with low recognition accuracy Character recognition with excellent workability The device can be realized.

[Brief description of drawings]

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

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

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

FIG. 4A is a diagram showing an example of a recognition target document in the present embodiment. FIG. 4B is a diagram showing a result of performing character pattern group classification processing on the recognition target document.

5 is a diagram showing a character recognition result of the recognition target document in FIG. 4;

FIG. 6A is a diagram showing a state in which a recognition result batch conversion process is being performed in this embodiment. FIG. 6B is a diagram showing the corrected recognition result.

FIG. 7 is a functional block diagram of a conventional character recognition device.

[Explanation of symbols]

 1 image input unit 2 image data storage unit 3 circumscribed rectangle detection unit 4 figure feature extraction unit 5 recognition dictionary 6 character recognition unit 7 recognition result storage unit 8 recognition result correction unit 9 recognition result output unit 10 character pattern group classification unit 11 character pattern Group classification result storage unit 12 Graphic feature extraction unit 13 Recognition result batch correction unit 14 Scanner 15 CPU 16 Keyboard 17 Read-only memory 18 Control program area 19 Recognition dictionary area 20 Random access memory 21 Image data area 22 Outer rectangle area 23 Character pattern group Area 24 Recognition result area 25 Display device 26 Internal bus

Claims (1)

[Claims] 1. An image input unit for inputting a document to be recognized and converting it into image data, an image data storage unit for storing the image data output from the image input unit, and an image data storage unit for storing the image data. A circumscribed rectangle detection unit that detects a rectangle circumscribing a character in image data based on a series of black pixels,
A character pattern group classification unit that classifies the character patterns in the circumscribed rectangle detected by the circumscribed rectangle detection unit into a character pattern group by a superposition method, and a character that stores the character pattern group classified by the character pattern group classification unit. A pattern group classification result storage unit, a graphic feature extraction unit that extracts the distribution of black pixels in the representative pattern of the character pattern group classified by the character pattern group classification unit as a graphic feature, and graphics of all recognition target characters in advance. A recognition dictionary that stores features,
A character recognition unit that obtains a recognition result by comparing the graphic feature extracted by the graphic feature extraction unit with the recognition dictionary, a recognition result storage unit that stores the recognition result obtained by the character recognition unit, and the recognition result storage unit A recognition result for inputting a correct recognition result for a character having a low recognition accuracy stored in and correcting all recognition results of the same character pattern group as that character stored in the character pattern group classification result storage section. A character recognition device comprising: a batch correction unit; and a recognition result output unit that outputs a corrected recognition result stored in the recognition result storage unit.