JPH11282958A - Character recognition method, its device and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform normal character recognition even when a character inside a frame is in contact with a frame ruled line. SOLUTION: A white picture element connection component area 102 inside a frame 100 is extracted, and black picture element connection component areas 104, 105 and 106 are extracted inside the above area. By checking the length of the white run of the outer periphery of the area 102, the contact of the character and the ruled line on the lower side of the frame 100 is estimated and the lower side of the area 106 is moved upwards by one or two picture elements for separation. Thereafter, the image of the area 104 and the image of the integrated area of the area 105 and the area 106 are segmented and the character is recognized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition technique, and more particularly to a technique for recognizing a character inside a frame surrounded by ruled lines such as a table or a form.

[0002]

2. Description of the Related Art Conventionally, when recognizing characters written in a frame surrounded by ruled lines such as a table or a form, Japanese Patent Laid-Open Publication No.
As described in Japanese Patent No. 172984, first, ruled line rectangles are extracted in the main scanning direction and the sub-scanning direction, and a frame surrounded by the ruled line rectangles is recognized using the coordinates of the ruled line rectangles. Next, the rectangle of the black pixel connected component is extracted inside the recognized frame, but the contact between the rectangle of the black pixel connected component and the frame is checked so as not to cut out part of the ruled line of the frame as a character by mistake. The rectangle in contact with the frame is removed, and character recognition is performed by cutting out a character image in the frame using only the rectangle of the black pixel connected component remaining without being removed.

[0003]

According to the above-mentioned prior art, if a character in a frame is in contact with a ruled line, a black pixel connected component rectangle of the character is removed.
There was a problem that character images could not be cut out correctly and could not be recognized.

Accordingly, an object of the present invention is to enable a character in a frame to be accurately recognized even when a character is in contact with a ruled line of the frame.

[0005]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention extracts a white pixel connected component region of a binary image such as a form, and forms a binary image of the binary image inside the extracted white pixel connected component region. A black pixel connected component region is extracted, and a process for separating the character and the ruled line is performed on the black pixel connected component region in the extracted black pixel connected component region where it is estimated that there is contact between the character and the ruled line. Thereafter, a character image is cut out from the binary image based on the information of the black pixel connected component area, and character recognition is performed. Further, by examining the length of the white run on the outer periphery of the white pixel connected component area, a black pixel connected component having a contact between a character and a ruled line is estimated.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A character recognition system using a computer will be described below as an embodiment of the character recognition method and apparatus of the present invention. FIG. 1 shows a processing flow of this character recognition system, FIG. 2 shows a functional block configuration, and FIG. 3 shows a hardware configuration.

The hardware configuration shown in FIG. 3 is a general computer, and reference numeral 50 denotes a CPU and 51 denotes a C
A RAM used as a work area of the PU 50, a storage area of programs and data, a hard disk drive 52,
53 is a display device, 54 is a pointing device such as a mouse or an input device such as a keyboard, 55 is an image scanner, 56 is a floppy disk or CD-R.
The drive is a drive for reading and writing a portable recording medium such as an OM, and these components are interconnected via a system bus 58.

In order to execute the processing consisting of steps S1 to S8 shown in FIG. 1 on the hardware of this computer, or a binary image input unit 20, a recognition area information input unit 21, and a white pixel connection unit shown in FIG. A character recognition processing program 68 for realizing the functions of the component extracting unit 22, the black pixel connected component extracting unit 23, the character / ruled line separating unit 24, the character image generating unit 25, and the character recognizing unit 26 is stored in the RAM 51 and executed by the CPU 50. Be executed. The character recognition processing program 68 is read from a portable recording medium 57 such as a floppy disk on which the program is recorded by the drive 56 and directly loaded into the RAM 51, or is temporarily stored in the hard disk device 52 and executed. Sometimes, it is loaded from the hard disk device 52 to the RAM 51. As various memories 30 to 36 in FIG.
Specific areas 60 to 66 of the AM 51 are allocated. R
A recognition dictionary 67 which is a set of standard patterns or templates for character recognition is loaded from the recording medium 57 or the hard disk device 52 into the AM 51. A control program such as an operating system is also loaded from the hard disk device 52 into the RAM 51.

The operation of the character recognition process will be described below mainly with reference to FIGS. First, in step S1, 2
A master image of a form to be processed (a binary image of an empty form with no data) is input by the value image input unit 20 and stored in the master image memory 30. Specifically, for example, an empty form is read by the image scanner 55, and the read image is stored in the master image memory 30, or a master image read in advance and stored in the hard disk device 52 is stored in the master image memory. Read in 30.

In the next step S2, the recognition area information input unit 21 causes a character recognition frame on the form (a rectangular area surrounded on all sides by a horizontal ruled line and a vertical ruled line).
The information (recognition area information) such as the position of the character and the character type to be recognized is input and stored in the area information memory 32. Specifically, for example, the master image input in the previous step S1 is displayed on the screen of the display device 53, and the input device 5
By operating the mouse 4 etc., two opposing vertices of the frame are designated on the screen, the coordinates are input, and the character type and the like are designated. If a form of the same type as a form that has already been processed is handled, and the recognition area information is stored in the hard disk device 52 as a file, the type of the form is designated and the recognition area information is specified. Recognition area information can also be input by a method of reading into the area information memory 32. In this case, there is no need to input the master image again (step S1 can be omitted).

In the next step S3, the binary image input unit 20
Thus, a data image of the same type of form (data to be subjected to character recognition, that is, a binary image of a form in which characters including numbers and symbols are entered) is input and stored in the data image memory 31. Specifically, for example, the form in which data is written is read by the image scanner 55, and the read image is stored in the data image memory 31. Alternatively, a data image previously read by the image scanner 55 and stored in the hard disk device 52 is read into the data image memory 31.

In the next step S4, the white pixel connected component extraction unit 22 refers to the position information of the frame to be recognized in the area information memory 32, scans the data image inside the frame, and connects the white pixels. The component is extracted, and the position information (for example, the coordinates of two opposing vertices) of the rectangular region surrounding the white pixel connected component is stored in the white connected component memory 33 as the information of the white pixel connected component region. When there are a plurality of frames to be recognized on the form, a similar white pixel connected component area is extracted for each of the frames. In this process, the white pixel connected component extraction unit 22 also detects the length of the white run on the outer periphery of the white pixel connected component area, and stores the white run length information in the white connected component memory 33.

A description will be given with reference to FIG. In FIG.
100 is a ruled line constituting a frame, and 102 is a white pixel connected component area to be extracted. This white pixel connected component area 1
02, the length of the horizontal white run on the upper side H1 and the lower side H2 and the length of the vertical white run on the left side V1 and the right side V2 are detected.

In the next step S5, the black pixel connected component extraction unit 23 refers to the information of the white pixel connected component area in the white connected component memory 33, and the data in the data image memory 31 for the white pixel connected component area. The image is scanned, the connected components of the black pixels are extracted, and the position information (for example, the coordinates of two opposing vertices) of the rectangular region surrounding the black pixel connected components is stored in the black connected component memory 34 as the black pixel connected component region information. . In the case of the frame shown in FIG. 4, the black pixel connected component region 104 corresponding to the character "A" written in the frame, the black pixel connected component region 105 corresponding to the upper portion of the character "U", and the lower portion correspond to the lower portion. The black pixel connected component area 106 is extracted. The extraction processing of the black pixel connected component area is performed on the white pixel connected areas of all the frames to be recognized.

In the next step S6, the character / ruled line separating section 24 refers to the information on the white pixel connected component area and the white run length in the white connected component memory 33 to determine whether or not the ruled line of the frame is in contact with the character. Judge and estimate the contact position. If there is no contact between the character and the ruled line, the length of the outer periphery (upper side, lower side, left side, right side) of the white pixel connected component area should be substantially equal to the length of the white run detected in each. However, if there is a contact between the ruled line and the character, the white run is divided at the contact portion, and the above-mentioned relationship is not established. Therefore, by checking the length of the white run on the outer periphery of the white pixel connected component area, it is possible to determine the presence or absence of contact between the ruled line of the frame and the character. When it is determined that there is a contact between a character and a ruled line, by checking the length of a white run, it is possible to estimate a black pixel connected component region where a contact between the character and a ruled line occurs.

In the example shown in FIG. 4, the white pixel connected component area 1
02, the length of the left side V1, the length of the right side V2, and the length of the white run detected there are almost equal. However, the lower side H
In FIG. 2, two white runs divided before and after the contact position between the character and the ruled line are detected, and the lengths w1 and w2 are significantly different from the length of the lower side H2. Are determined to be in contact. Since the white run is divided at the contact position, it is estimated that there is a contact between the character and the ruled line at the lower end of the black pixel connected component area 106. Such contact between the character and the ruled line occurs due to a displacement of the character entry position and also due to a local displacement of the ruled line.

When the character / ruled line separation unit 24 determines that there is a contact between the character and the ruled line, the character / ruled line separation unit 24 refers to the black pixel connected component area information in the black connected component memory 34 and there is a contact between the ruled line and the character. The black pixel connected component region estimated to be corrected to remove the pixels of the ruled line from the region as noise, that is, to separate the character and the ruled line,
The corresponding information in the black connected component memory 34 is rewritten with the corrected information of the black pixel connected component area. In the case of the example of FIG. 4, since it is estimated that the contact between the ruled line constituting the lower side of the frame and the character has occurred in the black pixel connected component area 106,
For example, the correction is made such that the lower side of the black pixel connected component area 106 is moved upward by one pixel or two pixels, for example. Thus, the character line and the ruled line below the character "U" are separated. For the black pixel connected component region estimated to be in contact with the ruled line on the upper side of the frame, the upper side is set to, for example, 1
Moved down by a pixel or two pixels, for a black pixel connected component area that is presumed to be in contact with the ruled line on the left or right side of the ruled line of the frame, the left side to the right, or the right side to the left, For example, by moving by one or two pixels, the ruled line and the character are separated.

In the next step S7, the character image generation unit 25
4, the image of the black pixel connected component area that is regarded as one character alone, such as the black pixel connected component area 104 in FIG. 4, or the image in FIG. Black pixel connected component area 105,1
A character image is generated by cutting out an image of an area obtained by integrating two or more black pixel connected component areas regarded as one character, such as 06, from the data image in the data image memory 31 and generating the character image. It is stored in the image memory 35. Even in an area where a character and a ruled line are in contact with each other, as in the black pixel connected component area 106 in FIG. 4, the area in the previous step S7 is modified so as to separate the contact. A separated character image is generated.

In the next step S8, the character recognition section 26 reads a character image from the character image memory 35 to extract features, and matches the extracted features with the recognition dictionary 67 (FIG. 3) to identify characters. The recognition is performed, and the recognition result is stored in the recognition result memory 36. At the time of this recognition, the character type is narrowed down by referring to the information on the character type to be recognized in the area information memory 32. Characters such as the character “U” in FIG. 4 that are in contact with the ruled line of the frame can be recognized with high accuracy because a character image is generated in a form separated from the ruled line. If there is the same type of form to be processed next, the processing from step S3 to step S8 is repeated to perform character recognition.

It is needless to say that the present invention can be implemented by dedicated hardware having a configuration as shown in FIG. 2 without using a computer.

[0021]

As described in detail above, according to the character recognition method and apparatus of the present invention, in a table or a form,
There is a great effect that accurate character recognition can be performed even when a character written in the frame is in contact with the ruled line of the frame.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a flow of a process according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a functional block configuration according to an embodiment of the present invention.

FIG. 3 is a block diagram illustrating a hardware configuration according to an embodiment of the present invention.

FIG. 4 is a schematic diagram for explaining processing contents.

[Explanation of symbols]

 Reference Signs List 20 Binary image input unit 21 Recognition area information input unit 22 White pixel connected component extraction unit 23 Black pixel connected component extraction unit 24 Character / ruled line separation unit 25 Character image generation unit 26 Character recognition unit 30 Master image memory 31 Data image memory 32 Area information memory 33 White connected component memory 34 Black connected component memory 35 Character image memory 36 Recognition result memory 50 CPU 51 RAM 52 Hard disk device 53 Display device 54 Input device 55 Image scanner 57 Portable recording medium 68 Character recognition processing program 100 frame Ruled line 102 White pixel connected component area 104, 105, 106 Black pixel connected component area

Claims (5)

[Claims] 1. A first step of extracting a white pixel connected component region of a binary image such as a form, and a black pixel connected component of the binary image inside the white pixel connected component region extracted by the first step A second step of extracting a region;
In the black pixel connected component area extracted by the step of
A process for separating a character and a ruled line from a black pixel connected component area estimated to be in contact with a character and a ruled line is performed.
Step, a fourth step of cutting out a character image from the binary image based on the information of the black pixel connected component area after the processing of the third step, and character recognition of the character image cut out by the fourth step. Performing the fifth step. 2. The character recognition method according to claim 1, wherein
In a third step, a character recognition method is characterized by estimating a black pixel connected component region where a character is in contact with a ruled line by examining the length of a white run on the outer periphery of the white pixel connected component region. A first means for storing a binary image such as a form; a second means for extracting a white pixel connected component area of the binary image stored in the first means; Third means for extracting a black pixel connected component area of a binary image stored in the first means inside the white pixel connected component area extracted by the means, and black pixels extracted by the third means Fourth means for performing processing for separating a character and a ruled line from a black pixel connected component area in which it is estimated that there is contact between a character and a ruled line in the connected component area, and black after processing by the fourth means. Fifth means for cutting out a character image from the binary image stored in the first means based on the information of the pixel connected component area, and sixth means for performing character recognition of the character image cut out by the fifth means And a character recognition device. 4. The character recognition device according to claim 3, wherein
A fourth means is a character recognition device for estimating a black pixel connected component region where a character contacts a ruled line by examining a length of a white run on an outer periphery of the white pixel connected component region. 5. The first, second, third, and fourth steps of the character recognition method according to claim 1 or 2.
A computer-readable recording medium on which is recorded a program for causing a computer to execute the steps of the first and second steps.