JP2000029982A - Character recognition device and character recognition result output method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make correcting efficient by extracting a character whose calculated conviction degree does not reach a specified value and outputting its recognition result and character image, side by side. SOLUTION: The character image is cut out of a binary image and stored in an image area of a RAM 5, and the sheet number of the sheet where the character image is present, the page number of the character in the same sheet, an image area, a start point, an end point, and an item are stored in a ROM 5. Character recognition is performed by referring to a dictionary in a storage device 6 and the result is stored as a recognition result. The feature quantity of the character image is extracted and compared with a standard feature quantity to calculate the degree of conviction, which is stored as a degree of conviction. The calculated conviction degree is compared with a threshold and in the case of 'conviction degree < threshold', a flag is set to '1' (ON). The input image and recognition result for which the flag is set ON are taken out of the RAM 5 by characters or items, sorted in the sorting order of the characters, and outputted to a display 8 or printer 9.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for performing character recognition on an input character image and outputting a character having a suspicious recognition result, and a method of outputting a recognition result.

[0002]

2. Description of the Related Art The recognition rate of character recognition is not necessarily 100%.
is not. In recent years, various technologies have been developed and the recognition rate has been improved. However, since various types of fonts and character types have been increased, it is impossible to achieve a recognition rate of 100% as a result. Therefore, it is important how to correct the misrecognized recognition result. Since the correction work is performed by a person, it is necessary to reduce the burden on the person as much as possible. As a method of outputting a recognition result for facilitating the correction, a technique of displaying both the obtained recognition result and an input image as a source on a screen (for example, Japanese Patent Publication No. 7-72903), a method of character recognition, and the like. Techniques for calculating the degree of certainty of the result and changing the display according to the result, or inserting special characters before and after the result to make the result suspicious (for example, Japanese Patent Laid-Open No. 5-1).
No. 28301).

Here, the certainty factor is a numerical value indicating how likely a certain result is. It is expressed by an integer value from 0 to 100, and the higher the numerical value, the more certain. The certainty factor of the recognition result for each character can be calculated by, for example, a method described in Japanese Patent Application Laid-Open No. Hei 4-218883.

[0004]

According to the above-described conventional character recognition apparatus, in a method of outputting a recognition result, characters that are correctly recognized and characters that are suspicious of the recognition result are displayed in a mixed manner. Unrecognized characters must be searched from among them, and the eyes move so much that it is difficult to perform efficient correction. In particular, when inputting characters written on multiple sheets at once, performing continuous recognition processing and correcting the recognition results collectively, the corrector searches for unrecognizable characters from a large amount of character recognition results. This is very inefficient, since suspicious characters with recognition results appear randomly at random.

SUMMARY OF THE INVENTION An object of the present invention is to provide a character recognition apparatus and a character recognition result output method which enable an efficient correction operation by improving the recognition result output method.

[0006]

According to the present invention, there is provided a character recognition apparatus for inputting a character image, performing character recognition of the character image, and calculating the certainty of the recognition result for each character. Means for extracting characters that do not reach the value,
Means for outputting a recognition result of each extracted character in parallel with the character image. Means for extracting a character whose calculated certainty does not reach a predetermined value and belonging to a predetermined item; and recognizing a result of recognition of each extracted character in a specified item including the character image. And a means for outputting all the character images in parallel with the character recognition device. Means for extracting a character whose calculated certainty does not reach a predetermined value and which belongs to a character group specified in advance as a character that is likely to be erroneously recognized as another character; and Means for outputting a recognition result in parallel with the character image. Also, the invention is characterized by a method of outputting a character recognition result having such means as steps. Furthermore, a storage medium of a program having such steps as processing steps of a computer is characterized.

[0007]

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

FIG. 1 is a configuration diagram of the character recognition device of the present embodiment. As shown in FIG. 1, the character recognition device includes a scanner 2 for optically reading an input image, an input device 1 such as a keyboard for correcting a recognized character, and a central processing unit 3 for controlling each unit. And the central processing unit 3
And a ROM 4 for storing various programs such as a character recognition process to be executed by the CPU, and an R for storing an input image and a table described later.
AM5, a storage device 6 such as a ROM or a RAM for storing a dictionary for registering standard features of characters, and a work area RAM for temporarily storing various data during character recognition processing
7 and a display 8 for displaying recognition characters with low confidence.
And a printing device 9, and these components are connected by a bus. The present invention can also be implemented by using a general-purpose computer such as a personal computer instead of the dedicated character recognition device, and executing a program for performing character recognition processing and the like stored in the main storage device. This program is stored in a storage medium, read into a main storage device of a computer via a drive device (not shown), and can be executed by the computer.

FIG. 2 is a diagram showing a data format of a table on the RAM 5. Each row of the table corresponds to each recognized character, the sheet number 31 is a number representing the number of the sheet, the item number 32 is a unique character number in the same sheet, and the image area 33 is input on the RAM 5. The start address of the area for storing the extracted character image, the recognition result 34 is the result of character recognition,
The starting point 35 is the upper left coordinate value of the area where the character exists on the sheet, the ending point 36 is the lower right coordinate value of the area, the item 37 is the number of the data item to which the character belongs, and the confidence 38 is the recognition result of the character. The flag 39 is a flag that is set when the certainty factor is less than the threshold value.

FIG. 3 is a flowchart showing the flow of processing of the character recognition processing program. The program reads the input data on the sheet via the scanner 2 and forms a binary image in step 21. Next, in step 22, a character image is cut out from the binary image and stored in the image area on the RAM 5, and the sheet number 31 of the sheet in which the character image exists, the item number 32 of the character in the same sheet, the image area 33, the starting point 35, an end point 36, and an item 37 are stored in the RAM 5. Next, in step 23, character recognition is performed with reference to the dictionary on the storage device 6, and the result is stored in the recognition result 34. Next, in step 24, the feature amount of the character image is extracted, and the extracted feature amount is compared with a standard feature amount registered in the dictionary to calculate a certainty factor. Store. Next, the predetermined threshold value is compared with the certainty factor calculated in step 24,
When "confidence <threshold" (Step 25: Y)
Proceeds to step 26, and sets "1" to the flag 39 (sets the flag). Conversely, if “confidence ≧ threshold” (step 25: N), no flag is set. When steps 21 to 26 are completed for the character images on all sheets (step 27: Y), the processing of step 28 is performed. In step 28, the input image with the flag set in step 26 and the recognition result 34 are taken out of the RAM 5 one character at a time or one item at a time, sorted according to the character sort order and output to the display 8, or Output to

FIG. 4 is a flowchart showing the processing of step 28 in an expanded manner, showing a first method of outputting an input image and a recognition result. In step 41, the recognition result 34 is read from the RAM 5 character by character. Read recognition result 34
It is determined in step 42 whether the flag 39 is set. When the flag is not set (Step 42: N)
Returns to step 41 and recognizes the next character as the recognition result 34
Is read. If the flag is set (step 42: Y), the recognition result and the input image are stored in the work area RAM 7 (step 43). Step 4
Steps 1 to 43 are completed for all characters (step 4
4: Y) If so, the process proceeds to step 45. In step 45, the characters of the recognition result 34 are sorted according to the sort order of the characters. Next, in step 46, the input image and the recognition result are
The characters are displayed on the display 8 one by one or output to the printing device 9.

FIG. 5 is a diagram showing a display example of an input image and a recognition result in the first output method. The input image 51 is displayed one character at a time, and the recognition result 52 for the character is displayed. The input image and the recognition result are displayed in parallel (53). The display order is the order in which the characters displayed in the recognition result 52 are the most. In FIG. 5, x (54) of recognition results “1”, y (55) of “5”, and “7” are recognized.
Are shown as z (56) (where x> y
> Z). 54, 55, so that the certainty is in ascending order.
It is also possible to rearrange them in 56. By clicking the input image 51 with a mouse or the like,
The program refers to the RAM 5 and displays a series of character images in the item to which the designated character belongs on the display 8. This makes it possible to know the image of the character and a series of characters existing before and after the character. Input device 1
When any of the recognition results 52 displayed via the is corrected, the apparatus corrects the corresponding recognition result 34 on the RAM 5.

FIG. 6 is a flow chart showing the processing of step 28 in an expanded manner, showing a second output method of the input image and the recognition result. In step 61, the number of the item to be confirmed and to be corrected is input via the input device 1. The item refers to a data item composed of a series of character strings as shown in FIG. These items are set with unique numbers in advance. The plurality of input sheets have the same data format having the same items. In step 62, the recognition result 34 is read from the RAM 5 character by character. It is determined in step 63 whether the flag 39 is set on the read recognition result 34. If the flag is not set (step 63: N), the process returns to step 62 to read the flag 39 of the recognition result 34 for the next character. If the flag is set (step 63: Y), it is determined in step 64 whether the character is a character in the item designated in step 61. If the character is outside the specified item (step 64:
N), returning to step 62; recognition result 3 for the next character
4 is read. If the character is within the designated item (step 64: Y), the recognition result and the input image are stored in the work area RAM 7 for each item (step 6).
5). If the processing of steps 62 to 65 is completed for all the characters (step 66: Y), the process proceeds to step 67. In step 67, the characters of the recognition result 34 are sorted according to the sort order of the characters. Next, in step 68, the input image and the recognition result are displayed for each item on the display 8 or output to the printing device 9.

FIG. 7 is a diagram showing a display example of an input image and a recognition result in the second output method. The input image 71 extracted for the specified item is displayed, and the recognition result 72 for the item is displayed. Here, not all the recognition results in the item are displayed, but only the recognition result with the flag set. The input image and the recognition result are displayed in parallel (73). The display order is the recognition result 72.
Are displayed in descending order of the characters displayed in. Also, the number 74 of the currently displayed item is displayed. When it is desired to display a recognition result of another item number, the input device 1 designates the item number. If any character in the recognition result 72 displayed via the input device 1 is corrected,
The device corrects the corresponding recognition result 34 on the RAM 5.
As described above, since an image of a series of character strings in the item to which the character displayed in the recognition result 72 belongs is displayed on the input image 71, this method is effective when this character string has some meaning.

FIG. 9 is a flowchart showing the processing of step 28 in an expanded manner, showing a third method of outputting an input image and a recognition result. In step 91, the item number of the pattern to be confirmed and corrected is input via the input device 1. A pattern refers to a group of characters that are likely to be erroneously recognized as other characters as shown in FIG. Unique item numbers and thresholds corresponding to each pattern are stored in the work area RAM in advance.
7 is stored. In step 92, the recognition result 34 is
Read one character at a time from M5. Read recognition result 34
It is determined in step 93 whether the flag 39 is set. When the flag is not set (Step 93: N)
Returns to step 92 and recognizes the next character as the recognition result 34
Is read. If the flag is set (step 93: Y), it is determined in step 94 whether the character belongs to the pattern specified in step 91. If it is a character other than the specified pattern (step 94: N),
Returning to step 92, the flag 39 of the recognition result 34 is read for the next character. If the character has the designated pattern (step 94: Y), the recognition result and the input image are stored in the work area RAM 7 for each character (step 95).
When the processing of steps 92 to 95 is completed for all the characters (step 96: Y), the process proceeds to step 97. In step 97, the characters of the recognition result 34 are sorted according to the sort order of the characters. Next, at step 98, the input image and the recognition result are displayed on the display 8 for each character, or output to the printing device 9.

FIG. 10 is a diagram showing a display example of an input image and a recognition result in the third output method. The input image and the corresponding recognition result are displayed one by one as in the case of FIG. Pattern item number 10 specified in step 91
1 is displayed. To display a result of another pattern, another pattern item number is designated by the input device 1. FIG. 10 is an output example when the pattern of item No. 1 in FIG. 11 is designated, where x (102) recognition results “1” and y are “7” recognition results
(103) (where x>
y). When any character in the recognition result displayed via the input device 1 is corrected, the device corrects the corresponding recognition result 34 on the RAM 5. Thus, while the first output method targets all characters for extraction, this method narrows the target to specific characters that are easily misrecognized. it can.

FIG. 12 is a flowchart showing the processing of step 28 in an expanded manner.
The output method of is shown. The user confirms via the input device 1, inputs the item number of the pattern to be corrected (step 121), and inputs the number of the target item (step 122). In step 123, the recognition result 34 is read from the RAM 5 character by character. It is determined in step 124 whether the read recognition result 34 has the flag 39 set. If the flag is not set (Step 124: N), Step 123
And the flag 39 of the recognition result 34 is read for the next character. If the flag is on (step 124:
Y) determines whether the character belongs to the pattern specified in step 121 in step 125. If it is a character other than the designated pattern (step 125: N), the process returns to step 123 to read the recognition result 34 for the next character. If the character has the specified pattern (step 125:
Y) In step 126, it is determined whether the character is the character of the item designated in step 122. If it is a character other than the designated item (N in step 126), the process returns to step 123 to read the next recognition result character. If the character of the specified item (step 1
26: Y), the recognition result and the input image are stored for each item in the work area RAM 7 (step 127). If the processing of steps 123 to 127 is completed for all characters (step 128: Y), the process proceeds to step 129. In step 129, the characters of the recognition result are sorted according to the sort order of the characters. Next, in step 130, the input image and the recognition result are displayed for each item on the display 8 or output to the printing device 9.

FIG. 13 is a diagram showing a display example of an input image and a recognition result in the fourth output method. The input image and the corresponding recognition result are displayed for the specified item as in the case of FIG. However, the displayed recognition result is limited to characters belonging to the specified pattern. The designated item number is 131 and the designated pattern item number is 13.
2 is displayed. Also in this case, the corresponding recognition result 34 on the RAM 5 can be corrected by correcting the character in the recognition result. According to the fourth output method, the effects of the second and third output methods can be obtained together.

[0019]

As described above, according to the present invention, only the recognition result of a character having a low degree of certainty is output for each character or for a specific item. Efficient correction work can be achieved without having to search for characters. In addition, by specifying a group of characters that are likely to be erroneously recognized, the characters to be corrected can be narrowed down, and an efficient correction operation can be performed.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a character recognition device according to an embodiment.

FIG. 2 is a diagram illustrating a data format of a table according to the embodiment.

FIG. 3 is a flowchart illustrating a processing flow of a character recognition processing program according to the embodiment;

FIG. 4 is a flowchart illustrating a flow of a first output process of an input image and a recognition result.

FIG. 5 is a diagram illustrating a display example of a first output process.

FIG. 6 is a flowchart illustrating a flow of a second output process of an input image and a recognition result.

FIG. 7 is a diagram illustrating a display example of a second output process.

FIG. 8 is a diagram showing an example of an item.

FIG. 9 is a flowchart illustrating a flow of a third output process of an input image and a recognition result.

FIG. 10 is a diagram illustrating a display example of a third output process.

FIG. 11 is a diagram illustrating an example of a pattern that is easily recognized erroneously and a threshold value thereof.

FIG. 12 is a flowchart illustrating a flow of a fourth output process of an input image and a recognition result.

FIG. 13 is a diagram illustrating a display example of a fourth output process.

[Explanation of symbols]

34: recognition result, 37: item, 38: certainty factor, 39: flag, 51: input image, 52: recognition result

Claims (5)

[Claims] 1. A character recognition apparatus for inputting a character image, performing character recognition on the character image, and calculating a certainty factor of a recognition result for each character, wherein a character whose calculated certainty factor does not reach a predetermined value is extracted. And a means for outputting the recognition result of each extracted character in parallel with the character image. 2. A character recognition apparatus for inputting a character image, performing character recognition on the character image, and calculating certainty of a recognition result for each character, wherein the calculated certainty does not reach a predetermined value. Means for extracting a character belonging to a pre-designated item, and means for outputting a recognition result of each extracted character in parallel with all the character images in the designated item including the character image A character recognition device characterized in that: 3. A character recognition apparatus for inputting a character image, performing character recognition on the character image, and calculating the certainty of the recognition result for each character, wherein the calculated certainty does not reach a predetermined value. Means for extracting characters belonging to a group of characters designated in advance as characters that are easily misrecognized with other characters, and means for outputting a recognition result of each extracted character in parallel with the character image. A character recognition device comprising: 4. A character recognition method for inputting a character image, performing character recognition on the character image, and calculating certainty of a recognition result for each character, wherein characters whose calculated certainty does not reach a predetermined value are extracted. And outputting the recognition result of each extracted character in parallel with the character image. 5. A program embodied on a computer-readable storage medium, wherein the program inputs a character image, performs character recognition on the character image, and calculates certainty of a recognition result for each character. The program includes the following steps: (a) extracting a character whose calculated certainty factor does not reach a predetermined value; and (b) outputting a recognition result of each extracted character in parallel with the character image. .