JP3239965B2 - Character recognition device - Google Patents

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,
In particular, the present invention relates to a character recognition device capable of recognizing handwritten characters entered in a form with a non-dropout color character frame.

In recent years, with the spread of information processing apparatuses such as computers, a technique of character recognition for reading handwritten or printed characters written on paper or the like has been required as an input means to these apparatuses. This character recognition device optically reads and recognizes handwritten characters and printed characters written at specific positions on a previously designed form.

[0003]

2. Description of the Related Art In general, an OCR form for character recognition requires a character frame in order to specify a character entry position. However, in the process of recognizing a character, the character frame is redundant and interferes with character recognition, and it is necessary to remove the character frame before performing the recognition process.

[0004] Conventionally, as a method for preventing recognition from becoming impossible due to the presence of the character frame, printing of the character frame with a special ink of light red or light blue has been performed. This method utilizes the fact that a character frame printed in a dropout color does not respond to a reading device, and only characters written with a pencil or a ballpoint pen are taken into the reading device during scanning.

[0005]

However, the printing of a character frame using a dropout color uses a special ink, so the cost is very high, and the dropout color generally has many colors that are difficult for humans to see. However, there is a disadvantage that it is difficult to fill out a form.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned drawbacks, and provides a character recognition device capable of accurately recognizing characters written in a character frame printed in a non-dropout color. The purpose is to do.

[0007]

FIG. 1 shows the principle of the present invention. The character recognition apparatus comprises an image input unit 8 for reading a form image and an inner periphery of the character frame 2 detected by the character frame detection unit 3. From all black pixels in contact with
Track black pixels adjacent to the black pixel and the starting end, a predetermined length continuous
A contour tracing unit 4 for obtaining contour forming pixels by deleting black pixels that do not continue, a contour direction calculating unit 5 for calculating the direction of the contour forming pixels obtained by the contour tracing unit 4, and a contour direction line as a sentence.
Extending in the character frame 2 direction, the extension line in the character frame 2 is set as a black-and-white boundary line, and the character frame 2 other than the area surrounded by the black-and-white boundary line is converted into white pixels, and the interference portion with the character frame 2 And a character part extracting unit 7 for extracting a character 6 part from the character string.

[0008]

The read image 1 read by the image input unit 8 includes a character frame 2 printed on a form in advance and a handwritten or printed character 6 to be recognized, and is determined in advance when designing the form. The position of the character frame 2 is detected by the character frame detection unit 3. The detection of character frame 2
For example, it can be easily performed by a known means such as a peripheral distribution.

The contour tracing unit 4 extracts a black pixel in contact with the inner periphery of the character frame 2 detected by the character frame detecting unit 3, and performs contour tracing toward the inside of the frame with the black pixel as a starting end. The number of contour tracking pixels in the contour tracking unit 4 is, for example, 4 pixels.
Alternatively, black pixels less than this threshold value are set in advance as five pixels, and are regarded as character frame 2 and converted to white pixels.

The contour forming pixels tracked by the contour tracking unit 4 are a set of black pixels that are in contact with the character frame 2 and have a continuity of a predetermined length, interfere with the character frame 2, and
It is presumed that it constitutes a part of the character 6.

The contour direction line computing section 5 is provided with the contour tracing section 4.
Calculates the direction of the contour line from the contour forming pixels tracked in. The character portion extraction unit 7 extends the outline direction line obtained as described above in the direction of the character frame 2 and obtains a black and white boundary line in the character frame 2. The black-and-white border, which is an extension of the contour direction line into the character frame 2, is considered to be an extension of the contour of the character 6, and when black pixels other than the area surrounded by the black-and-white border are converted into white pixels, the character The character 6 component that interferes with the frame 2 can be extracted.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below in detail with reference to the accompanying drawings. First, as shown in FIG. 1, the form has a character frame 2 printed in a non-dropout color, and the characters 6 entered in the character frame 2 are taken in from the image input unit 8.

The image input unit 8 converts the form into image data,
Reference is made to form data (not shown) in which data such as a character entry position is stored, and an image of that part is input. The read image 1 captured by the image input unit 8 includes a character 6 and a character frame 2 component.

FIG. 2 shows an example of the read image 1. In the illustrated example, “8” is handwritten in the character frame 2 and the character 6
Shows an example in which the character is in contact with the character frame 2. As shown in FIG. 6, even if the character frame 2 is simply deleted or only the inside of the character frame 2 is to be recognized, erroneous recognition or recognition failure occurs. The case is shown. In FIGS. 2 and 6 illustrating the read image 1, black pixels are indicated by black circles, and white pixels are indicated by white circles.

Hereinafter, the recognition operation of the character 6 will be described by taking the above-described read image 1 as an example. As shown in FIG.
Prior to character recognition, the position of the character frame 2 is specified by the character frame detection unit 3.

To specify the frame position, a well-known marginal distribution method can be applied. FIG. 4 shows a detailed procedure for specifying the character frame 2 based on the marginal distribution. First, the number of black pixels in the read image 1 is counted in the vertical and horizontal directions, and the frequency distribution is obtained. Next, the peak of the histogram value obtained as described above is obtained, and a portion close to the edge of the read image and having a certain distribution height is specified as a frame position. Here, what is significant is the inside of the character frame 2, and its position is indicated by reference numerals a1 to a4 in FIG.
Is denoted by reference numeral b.

Thereafter, the control is transferred to the contour tracing unit 4 to trace the character contour. That is, as shown in FIG. 4, points inside the character frame 2 obtained by the above-described procedure are sequentially looked at, and if there is a black pixel, the outline of the black pixel is traced inward from the black pixel. .

The tracking of black pixels is performed until the number of pixels reaches a threshold value (5 pixels in this example). For example, a black pixel indicated by reference numeral c in FIG. Since there is no pixel, the tracking is terminated, and it is regarded as a part of the character frame 2 and is deleted.

Attention is now paid to the black pixel at d1, and the tracing of the contour is d2, d3, d4 and d5. This can be easily performed by looking at the pixel adjacent to the target black pixel. In this example, the threshold value is set to 5 pixels, but the effect of this processing can be obtained even with about 3 pixels, and this can be varied depending on the size of the target character frame 2.

The direction line calculating section 5 obtains a contour direction line from the contour forming pixels obtained by the contour tracking section 4. Here, the approximation is performed by linear approximation, which can be performed by a known technique such as the least square method. The contour direction lines obtained as described above are indicated by e1 to e10.

The character portion extracting section 7 is provided with the outline direction lines e1 to e1.
From 0, character constituent pixels that interfere with the character frame 2 are separated. That is, as shown in FIG.
1 to e10 are extended to the outside of the character frame 2, and a pixel position on the contour direction line is set as a separation boundary between the character 6 and the character frame 2.

In the example of FIG. 2, the point d1 is a boundary in which the upper part is white and the lower part is black because the upper part is a white pixel and the lower part is a black pixel. This process is sequentially performed on the black pixels that are in contact with the character frame 2 and at the transition point between white and black (or black and white).

From the black-and-white border line obtained as described above, erasing pixels determined to be white outside the border is as follows.
After completion of the processing, the image is as shown in FIG. 5, and there is no missing part of the character 6, no extra part, and it is recognized as "8" by the conventional recognition technology. It becomes possible.

Thereafter, the character 6 image is subjected to pattern matching with the recognition dictionary 10 in the recognition processing section 9 to recognize the character 6.

[0025]

As is clear from the above description, according to the present invention, in a form in which a character frame is printed in a non-dropout color, even if the character and the character frame touch or intersect, only the character Can be accurately cut out.

[Brief description of the drawings]

FIG. 1 is a principle diagram of the present invention.

FIG. 2 is an explanatory diagram showing an embodiment of the present invention.

FIG. 3 is a flowchart showing the operation of the present invention.

FIG. 4 is a flowchart illustrating operations of a character frame detection unit and an outline tracking unit.

FIG. 5 is a diagram showing a cut-out state of a character.

FIG. 6 is a diagram illustrating a cut-out state of a character.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Read image 2 Character frame 3 Character frame detection part 4 Outline tracking part 5 Direction line calculation part 6 Character 7 Character part extraction part

Claims (2)

(57) [Claims] 1. A character frame detecting unit for detecting a character frame from a read image, and a black pixel adjacent to an inner periphery of the detected character frame is tracked as a starting point to track an adjacent black pixel toward the inside of the character frame. Long continuous
A contour tracing unit that obtains a contour forming pixel by removing black pixels that are not used , a direction line calculating unit that calculates a contour direction line from a continuous direction of the contour forming pixels , In
A character recognition device comprising: a character portion extracting unit that converts a character frame other than an area surrounded by the black and white boundary line into a white pixel and extracts a character portion from an interference portion with the character frame, using the extension line as a black and white boundary line. . 2. The character recognition device according to claim 1, wherein said character frame detecting section detects a character frame from a peripheral distribution of the number of pixels.