JP2002230481A - Optical character reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an OCR(optical character recognition) which correctly recognizes a character superposed on a character frame. SOLUTION: An image of a form 1, read by an image acquisition part 10, is stored in an image memory 20, and thereafter, the image is segmented by a character area segmenting part 31 and the character frame in the image is detected by a frame line detecting part 32. The character frame is removed from the image segmented by the character area segmenting part 31, to obtain a character pattern by a frame line removal part 33. The character pattern is given to a character line width calculating part 35, and the width of lines constituting the character. It is determined by a character line complementing part 36 whether the character has been partially removed together with the character frame or not, and the character is complemented with lines of the width calculated by the character line width calculation part 35, when it has been removed. The complemented character pattern is given to a character recognition part 50 and is converted into a character code, on the basis of a recognition dictionary 60.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for extracting characters in an optical character reader (hereinafter, referred to as "OCR").

[0002]

2. Description of the Related Art FIG. 2 is a block diagram showing an example of a conventional OCR. The OCR includes an image acquisition unit 10 that optically reads an image of the form 1 on which characters to be read are described, by decomposing the image into pixels and optically reading the image. The image acquisition unit 10 is connected to an image memory 20 that temporarily stores the read form image. The character extracting unit 30 is connected to the image memory 20.

The character extracting section 30 includes a character area extracting section 31,
It comprises a frame line detection unit 32, a frame line removal unit 33, and a character extraction unit 34.

A character area extracting section 31 extracts an image of a character area from the image memory 20 based on information of a form format registering section 40 in which forms such as dimensions, line positions, number of characters, and character types of the form 1 are registered. Things. Frame line detector 3
Numeral 2 detects a frame line in the image of the character area.

The frame line removing unit 33 removes the frame line detected by the frame line detecting unit 32 from the image cut out by the character area cutting unit 31, and extracts only the character pattern. The character extracting unit 34 extracts individual character patterns in the character area, and outputs a character recognizing unit 50 to the output side.
Is connected.

The character recognizing unit 50 extracts the characteristics of the character pattern extracted by the character extracting unit 30, compares the characteristic with the standard pattern registered in the recognition dictionary 60, and outputs the most similar character code. Is what you do. Recognition dictionary 60
Is provided with information such as the character type of the form 1 from the form format registration unit 40. Further, the character recognition unit 50
The output side is connected to an operation unit 70, where knowledge processing such as word matching and confirmation /
Processing such as correction is performed.

FIGS. 3A to 3D are explanatory diagrams showing an example of processing by the OCR of FIG. Hereinafter, FIG.
The operation of FIG. 2 will be described with reference to (a) to (d). The form 1 to be read is read by the image acquisition unit 10 and a form image as shown in FIG.

Next, the character area extracting section 31 is activated, and the image memory 2 is stored on the basis of the information of the form format registering section 40.
From 0, as shown in FIG. 3B, an image of a character string including a character string to be read and a frame line for writing characters is cut out. Further, the frame line detection unit 32 is activated, the frequency distribution of black pixels projected in the horizontal direction and the vertical direction is calculated, and vertical and horizontal frame lines are detected. The image of the character area cut out by the character area cutout unit 31 and the information of the frame detected by the frame detection unit 32 are stored in a frame removal unit 33.
Given to.

In the frame line removing section 33, as shown in FIG. 3C, a frame line portion is removed from the image of the character area.
The character pattern obtained by removing the frame lines by the frame line removing unit 33 is provided to the character extracting unit 34.

In the character extracting section 34, as shown in FIG. 3D, individual character patterns in the character area obtained by removing the frame lines are extracted and output to the character recognizing section 50.

The character recognition unit 50 extracts the characteristics of each character pattern, and compares the characteristics with the standard patterns registered in the recognition dictionary 60. Then, the character code of the most similar standard pattern is output as the recognition result.

[0012]

However, the conventional OCR has the following problems. As shown in FIG. 3A, if the character (the numeral “2” in this figure) entered in the form 1 is over the frame line, the character cutout unit 30 performs a frame removal process. A part of the character is removed together with the line, and the character is divided as shown in FIG. For this reason, there is a problem that the character recognition unit 50 does not recognize the character correctly, resulting in erroneous reading or non-reading.

The present invention solves the problems of the prior art and provides an OCR capable of correctly cutting out and recognizing a character pattern overlapping a frame line.

[0014]

In order to solve the above-mentioned problems, the present invention provides an OCR system in which an image of a form on which a character frame for character entry and a character to be read are described is decomposed into pixels, and the OCR is performed. An image acquisition unit to be read, a frame line removal unit that removes the character frame from the image of the form read by the image acquisition unit to generate a character pattern, and a character line that calculates the line width of characters in the character pattern A width calculating unit, determining whether a part of the characters in the character pattern is removed together with the character frame, and when the character is removed, the removed portion is calculated by the character line width calculating unit. And a character recognition unit that recognizes the character pattern complemented by the character line complement unit based on a recognition dictionary and converts the character pattern into a character code.

According to the present invention, since the OCR is configured as described above, the following operation is performed. The image of the form read by the image acquisition unit is provided to the frame line removal unit, and the character frame portion is removed to generate a character pattern. The character line width of the character pattern is calculated by the character line width calculation unit. Further, the character pattern is given to the character line complementing section, and it is determined whether or not a part of the character is removed together with the character frame. When a part of the character has been removed, the removed portion is complemented by the line having the line width calculated by the character line width calculation unit. The complemented character pattern is provided from the character line complementing unit to the character recognizing unit, and is converted into a character code based on the recognition dictionary.

[0016]

FIG. 1 is a block diagram of an OCR according to an embodiment of the present invention. Elements common to those in FIG. 2 are denoted by the same reference numerals.

This OCR has an image acquisition unit 10 that optically separates the image of the form 1 on which the character to be read is described into pixels as in FIG. The image acquisition unit 10 is connected to an image memory 20 that temporarily stores the read form image. The image memory 20 is connected to a character extracting section 30A different from that shown in FIG.

The character extraction section 30A includes a character line width calculation section 35 and a character line complementation section 36 in addition to a character area extraction section 31, a frame line detection section 32 and a frame line removal section 33 similar to those shown in FIG. It has been added.

That is, the character area extracting section 31 is configured to store the form, such as the size, line position, number of characters, and character type of the form 1 to be read, based on information of the form format registering section 40 in advance.
The image of the character area is cut out from the image memory 20. The frame line detection unit 32 detects a frame line in the image of the character area. Also, the frame line removing unit 33
Removes the frame line detected by the frame line detection unit 32 from the image of the character region cut out by the character region cutout unit 31,
Only character patterns are extracted.

On the other hand, the character line width calculating section 35 calculates the width (number of pixels) of a line constituting a character based on the character pattern output from the frame line removing section 33. Also,
The character line complementing unit 36 complements the character line by connecting the character patterns separated by the frame line removing process by the frame line removing unit 33 with the line having the width calculated by the character line width calculating unit 35. This is for reproducing a normal character pattern. The reproduced character patterns are extracted as individual character patterns by the character line complementing unit 36 and are provided to the character recognizing unit 50.

The character recognizing unit 50 extracts the characteristics of the character pattern extracted by the character extracting unit 30A, compares the characteristic with a standard pattern registered in advance in the recognition dictionary 60,
It outputs the character code closest to the feature. The recognition dictionary 60 is provided with information such as the character type of the form 1 to be read from the form format registration unit 40. Further, an operation unit 70 is connected to the output side of the character recognition unit 50. The operation unit 70 performs knowledge processing such as word collation and processing such as confirmation / correction by an operator.

FIGS. 4A and 4B are explanatory diagrams showing an example of the operation of the character extracting section 30 in FIG. Hereinafter, the operation of FIG. 1 will be described with reference to FIGS. 4 (a) and 4 (b).

The form 1 to be read is read by the image acquisition unit 10 and the form image is stored in the image memory 20.
Is stored in Next, the character area extracting unit 31 of the character extracting unit 30A is activated, and based on the information of the form format registration unit 40, a character string to be read and a frame line for writing characters from the image memory 20. The image of the character area containing is cut out. Further, the frame line detection unit 32 is activated,
The frequency distribution of black pixels projected in the horizontal and vertical directions is calculated, and vertical and horizontal frame lines are detected. The image of the character area cut out by the character area cutout unit 31 and the information on the frame detected by the frame detection unit 32 are provided to a frame removal unit 33.

In the frame line removing section 33, the frame portion is removed from the image of the character area, and as shown in FIG.
5 given.

The character line width calculator 35 calculates the character line width W for each character as follows. First, black pixels constituting a character are counted, and the total number is set to A. Next, FIG.
The entire character pattern is scanned by a window having a size of 2 pixels × 2 pixels as shown in the upper left of FIG. At that time, 4 in the window
The number of positions where all points are black pixels is counted, and the number is set as Q. Next, the character line width W is calculated by the following equation (1). W = A / (A-Q) (1)

The character line width W calculated by the character line width calculator 35 is given to a character line complementer 36.

The character line complementing section 36 determines whether or not a part of the character line has been deleted together with the frame line based on whether or not black pixels exist on both sides of the removed frame line. If there are black pixels on both sides, the space between them is complemented with black pixels based on the character line width W given from the character line width calculation unit 35.
A star-shaped symbol in FIG. 4B indicates a black pixel complemented by a missing portion of the character pattern.

The individual character patterns in the character area obtained by complementing by the character line complementing unit 36 are output to the character recognizing unit 50. The character recognition unit 50 extracts the characteristics of each character pattern, compares the characteristics with a standard pattern registered in advance in the recognition dictionary 60, and outputs a character code closest to the characteristic.

As described above, the OCR according to the present embodiment employs the character line width calculation unit 35 for calculating the character line width W for each character pattern.
And a character line complementing unit 36 for complementing a character pattern lost in the frame line removal processing based on the calculated character line width W. This makes it possible to reproduce an almost complete character pattern from the character pattern overlapping the frame line, and has the advantage that the recognition rate can be improved.

The present invention is not limited to the above embodiment, and various modifications are possible. For example, there are the following modifications.

(A) The method of calculating the character line width W is not limited to the equation (1). For example, the calculation may be performed based on the distribution of the number of black pixels continuous in the vertical and horizontal directions, the ratio of black pixels in the character pattern, and the like.

(B) The number of missing portions of the character pattern complemented by the frame line complementing section 36 is not limited to one. For example, if the number "0" overlaps the horizontal frame line, there are two missing portions. In such a case, the black pixels may be supplemented so that the ends of the nearest black pixels are connected to each other.

[0033]

As described above in detail, according to the present invention, a character line width calculating unit for calculating a line width of a character from a character pattern, and a character pattern width removed by a frame line removing unit along with a character frame. There is a character line complementing section for complementing the missing portion with a line having the line width calculated by the character line width calculating section. This makes it possible to reproduce an almost complete character pattern from the character pattern overlapping the frame line, and improve the recognition rate.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an OCR showing an embodiment of the present invention.

FIG. 2 is a configuration diagram illustrating an example of a conventional OCR.

FIG. 3 is an explanatory diagram showing an example of processing by the OCR in FIG. 2;

FIG. 4 is an explanatory diagram showing an example of an operation of a character cutout unit 30A in FIG.

[Explanation of symbols]

 1 Form 10 Image Acquisition Unit 20 Image Memory 30 Character Extraction Unit 31 Character Area Extraction Unit 32 Frame Line Detection Unit 33 Frame Line Removal Unit 35 Character Line Width Calculation Unit 36 Character Line Completion Unit 40 Form Format Registration Unit 50 Character Recognition Unit 60 recognition dictionary 70 operation unit

Claims (1)

[Claims] An image acquisition unit that decomposes an image of a form in which a character frame for character entry and a character to be read are described into pixels and optically reads the image; and A frame line removal unit that generates a character pattern by removing a character frame, a character line width calculation unit that calculates a line width of a character in the character pattern, and a part of the character in the character pattern is together with the character frame. A character line complementer that determines whether or not the character line has been removed, and when the character line has been removed, complements the removed portion with a line having the line width calculated by the character line width calculator; An optical character reading device, comprising: a character recognition unit that recognizes the character pattern complemented by (1) based on a recognition dictionary and converts the character pattern into a character code.