JP2000132636A - Character recognition method and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make precisely detectable a character frame position and to make recognizable a character filled in a character frame even if the character frame in image data is thinned or a background printing exists around the character frame by detecting the position of the character frame based on a frame area direction code obtained in the detection area of the character frame. SOLUTION: An area detecting a character frame is decided from document image data stored in an image data storage part 12 and character registration data containing a frame coordinate value stored in a character frame registration data storage part 13. A frame area direction code is detected with an object where black picture elements in the frame detection area are connected as an outline trace. The corner position candidate coordinate of the character frame is detected and is stored in a frame corner candidate coordinate storage part 18 against a picture where a candidate frame line generated based on a frame line candidate position is written. The position of a character frame inner-side is decided from the corner position candidate coordinate of the character frame stored in the frame corner candidate coordinate storage part 18 and character frame registration data stored in the character frame registration data storage part 13.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method and apparatus, and more particularly, to a character recognition method and apparatus for optically reading characters written in a character frame.

[0002]

2. Description of the Related Art When a character written in a character frame of a non-dropout color is recognized by a conventional character recognition device, black pixels are projected in a horizontal direction and a vertical direction in an area where a character frame is presumed to exist. Then, the histogram was obtained, and the position of the character frame was detected.

The estimated area is determined in the form image data in consideration of the printing displacement of the form paper in the vertical and horizontal directions, the geometric deformation of the image due to the error of the image scanner mechanism, and the like. The position of the character frame is detected from a position having a value greater than or equal to a threshold value obtained from the height and width of the character frame registered in advance on the histogram.

[0004]

However, for example, if the character frame is blurred in the form image data,
The histogram value at the character frame position becomes smaller than the threshold, and the character frame cannot be normally detected.

FIGS. 13A and 13B show a state in which a character frame cannot be detected normally. FIG. 13A is an explanatory diagram of form image data in which the character frame is blurred, and FIG. FIG. 9C is an explanatory diagram of a histogram based on pixel projection, and FIG. 9C is an explanatory diagram of a character frame position that is erroneously detected.

As shown in FIG. 13, a form image data 1 shows a character 3 entered in a character frame 2 (see (a)). From the histogram obtained by projecting pixels from the area 4 of the form image data 1 in the vertical direction, a threshold B derived from the height of the character frame 2 of the frame registration data
When a position that takes a histogram value equal to or greater than h is a frame position candidate, the positions L2, L5, and L10 of the faint character frame 2 are:
It is not selected because it takes a histogram value equal to or smaller than the threshold value Bh (see (b)). Instead, the position L1, which exceeds the threshold Bh and is close to the width of the character frame 2 and the pitch between the frames in the frame registration data,
L3, L4, L8, L9, and L13 are selected.

As a result of such selection, a frame F1 based on the positions L1 and L3, a frame F2 based on the positions L4 and L8, a position L9,
The frame F3 based on L13 is detected as the character frame 2 (see (c)), and the position of the character frame 2 is erroneously detected.

That is, in the conventional character recognition device, the position of the character frame 2 is detected by selection based on the histogram indicating the pixel distribution. Therefore, the form is determined by the characteristics of the image scanner and the reflectance of the character frame 2 printed on the form paper. When the frame line is blurred in the image data or when the background is printed around the character frame 2, the character frame 2 and the character 3 cannot be distinguished from the background, and the position of the character frame 2 cannot be accurately detected. .

In order to detect the character frame position by such selection based on the histogram indicating the pixel distribution,
For example, there is an image clipping device and a character recognition device disclosed in Japanese Patent No. 2761467.

[0010] An object of the present invention is to provide an image processing apparatus that can be used even if a character frame is blurred in image data or a background print is present around the character frame.
An object of the present invention is to provide a character recognition method and apparatus capable of accurately detecting a character frame position and recognizing characters written in the character frame.

[0011]

In order to achieve the above object, a character recognition method according to the present invention is a character recognition method for optically reading a character entered in a character frame and recognizing an image. The position of the character frame is detected based on a frame region direction code obtained in the region.

With the above-described structure, when a character written in a character frame is optically read and image-recognized, the position of the character frame is determined based on the frame region direction code obtained in the detection region of the character frame. Is detected. Thus, even if the character frame is faint in the image data or there is a background print around the character frame, the character frame position can be accurately detected and the character written in the character frame can be recognized.

Further, according to the character recognition device of the present invention,
The above character recognition method can be realized.

[0014]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a character recognition device according to an embodiment of the present invention. As shown in FIG. 1, the character recognition device 10 includes an image scanner unit 11, an image data storage unit 12, a character frame registration data storage unit 1.
3, a character frame position detection unit 14, a character recognition unit 15, a direction code storage unit 16, a frame candidate storage unit 17, and a frame corner candidate coordinate storage unit 18.

The image scanner section 11 optically reads a form in which characters are written in a character frame printed in a non-dropout color, photoelectrically converts the form, and acquires form image data composed of binary data. The acquired form image data is output to the image data storage unit 12. The image data storage unit 12 is a memory that stores the form image data read by the image scanner unit 11 as a two-dimensional image.

The character frame registration data storage section 13 is for storing character frame registration data including frame coordinate values registered in advance, and is composed of a magnetic disk, an optical disk, a floppy disk, or the like. The character frame registration data storage 13 stores a reference position and size of each character frame.

FIG. 2 is an explanatory diagram showing an example of character frame registration data stored in the character frame registration data storage. FIG.
As shown in the figure, the character frame registration data includes three character frames 19.
a, 19b, and 19c. The coordinates (X
_S, Y_S) are the upper left coordinates of the first frame on the form and the distance D
x1 and Dx2 are the distances from the first frame to the upper left of the second and third frames, the widths W1, W2, and W3 are the widths of each character frame and the height H
1, H2 and H3 indicate the height of each character frame, respectively.

The character frame position detection unit 14 compares the form image data with the character frame registration data by inputting the form image data from the image data storage unit 12 and the character frame registration data from the character frame registration data storage unit 13. To determine the position of the character frame. The image of the determined character frame position is supplied to the character recognition unit 15.

The character recognition unit 15 performs character recognition based on the form image data in the character frame from the input image of the character frame position. The direction code storage unit 16, the frame candidate storage unit 17, and the frame corner candidate coordinate storage unit 18 store the frame region direction code, the frame line candidate, and the frame corner candidate obtained when the character frame position detection unit 14 determines the character frame position. This is a memory for storing each data.

FIG. 3 is a flowchart showing the flow of a character frame position detecting process by the character frame position detecting section of FIG.
As shown in FIG. 3, first, a character frame is detected from the form image data stored in the image data storage unit 12 and the character frame registration data including the frame coordinate values stored in the character frame registration data storage unit 13. A character frame area determination process for determining an area is performed (step S101).

Next, the object formed by connecting the black pixels in the frame detection area obtained by the frame area determination processing is contour-traced to extract a frame area direction code, and the direction code is made to correspond to the outline coordinates. A frame area direction code extraction process to be stored in the storage unit 16 is performed (step S102).

Next, a candidate for a position inside the character frame is obtained from the frame area direction code stored in the direction code storage unit 16 and a frame line candidate detection process for storing the candidate in the frame line candidate storage unit 17 is performed (step S103).

Next, labeling is performed on the image in which the candidate frame lines created based on the frame line candidate positions stored in the frame line candidate storage unit 17 are written, and then the corner position candidate coordinates of the character frame are detected. Then, a frame corner candidate creation process to be stored in the frame corner candidate coordinate storage unit 18 is performed (step S10).
4).

Lastly, the character frame corner position candidate coordinates stored in the frame corner candidate coordinate storage unit 18 and the character frame registration data including the frame coordinate values stored in the character frame registration data storage unit 13 are used to determine the character frame. A frame position determination process for determining an inner position is performed (step S105).

As described above, the character frame position detecting process is performed by the character frame position detecting unit 14. The processing contents in each of the above-described steps will be described in detail below with reference to FIGS. FIG. 4 is an explanatory diagram showing an example of the form image data read by the image scanner unit of FIG. 1 and stored in the image data storage unit. 5 to 12 are explanatory diagrams of the character frame position detection operation from the form image data by the character frame position detection unit in FIG.

As shown in FIG. 4, the form image data stored in the image data storage unit 12 has three character frames 19a, 19b and 19c in which the character frame lines are blurred.
Cannot be detected normally.

The processing contents in the character frame area determination processing (step S101) will be described with reference to FIGS.

The position of the frame (character frame) in the form image data is shifted from the position registered in the frame registration data due to the mechanism accuracy of the image scanner 11 and the vertical and horizontal errors of the form paper. The geometric error depends on the mechanism accuracy and the form paper quality described above. here,
The error value of the form image data in the horizontal direction is N_H, the allowable error value in the vertical direction is N_V, and the detection area of each frame is +/- N in the horizontal direction with respect to the position indicated by the character frame registration data.
_H, and +/- N_V in the vertical direction.

When the upper left coordinate position of the nth character frame obtained from the character frame registration data is (Xnl, Ynu) and the lower right coordinate position is (Xnr, Ynd), a frame is formed for each frame. The upper, lower, left and right line segments are searched for in the upper left coordinate of the frame left vertical line detection area: (Xnl-N_H, Ynu-N_V) Lower right coordinate: (Xnl + N_H, Ynd + N_V) Upper left coordinate: (Xnr-N_H, Ynu-N_V) Lower right coordinate: (Xnr + N_H, Ynd + N_V) Upper left coordinate of the frame detection area of the horizontal line on the frame: (Xnl-N_H, Ynu-N_V) Lower right coordinate: (Xnr + N_H, Ynu + N_V) Upper left coordinate of the frame detection area of the horizontal line below the frame: (Xnl-N_H, Ynd-N_V) Lower right coordinate: (Xnr + N_H, Ynd + N_V)

FIG. 5A is an explanatory diagram showing two vertical line frame detection areas in the first frame of the form image data, and FIG.
Is an explanatory diagram showing two vertical line frame detection areas in the second frame of the form image data. FIG. 5C shows two vertical line frame detection areas in the third frame of the form image data. FIG. FIG. 6A is an explanatory diagram illustrating a frame detection area of two horizontal lines in the first frame of the form image data, and FIG. 6B is a diagram illustrating detection of two horizontal lines in the second frame of the form image data. FIG. 6C is an explanatory diagram showing a frame detection region of two horizontal lines in a third frame of the form image data.

The contents of the frame area direction code extraction processing (step S102) will be described with reference to FIGS.

FIG. 7A shows that the current trace point is i in the contour trace and the previous trace point is (i-
It is explanatory drawing which shows the frame area direction code in case of 1). In the contour trace, the search for the next trace point is performed counterclockwise around the current trace point. FIG. 7B is an explanatory diagram showing a frame area direction code when the outer peripheral contour of an object of 3 × 3 pixels is traced as an example. FIG. 7C is an explanatory diagram showing the data of the frame area direction code stored in the direction code storage unit.

FIG. 8A is an explanatory diagram showing a frame region direction code for a hole region having a size of 2 × 2 in an object. FIG. 8B is an explanatory diagram showing data obtained by the frame area direction code of FIG. 8A.

Next frame line candidate detection processing (step S10)
In 3), a frame region direction code 0 is detected on the left side of the line, a frame region direction code 4 is detected on the right side of the line, and a frame region direction code 6 is detected on the upper side of the line. Then, the lower side of the line has a feature that the frame area direction code 2 is detected.

As a result, each frame detection area shown in FIGS. 5 (a), 5 (b), 5 (c), 6 (a), 6 (b) and 6 (c) is obtained. Find the frame area direction code on the outer circumference of the object and the frame area direction code on the inner circumference when there is a hole,
It is stored in the direction code storage unit 16 in association with the contour coordinates.

The contents of the frame line candidate detection processing (step S103) will be described with reference to FIG.

In this process, the frame area direction code is read out from the direction code storage unit 16 and the outline pixel having the frame area direction code 4 which is a feature on the right side of the line is detected in the vertical line detection area on the left side of the frame. Direction, and the histogram (Hi
stgram). Similarly, in a vertical line frame detection area on the right side of the frame, pixels having a frame area direction code 0, which is a feature on the left side of the line, are projected in the vertical direction to generate a histogram.

Further, a contour pixel having a frame area direction code 2, which is a characteristic of the lower side of the line, in the upper horizontal line detection area of the same frame is projected in the horizontal direction to generate a histogram. Regarding the lower horizontal line of the frame, a contour pixel having a frame area direction code 6 which is a feature above the line in the detection area of the line is projected in the horizontal direction to generate a histogram.

After each process, the position where the second highest histogram value is detected in the order of the histogram value in each frame detection area is stored in the frame line candidate storage unit 17.

FIG. 9A is an explanatory diagram showing an example of a histogram obtained from the first frame. FIG. 9B is an explanatory diagram illustrating an example of a histogram obtained from the second frame. FIG.
(C) is an explanatory view showing an example of a histogram obtained from the third frame. In the above figures, the frame line candidate position is X1
The X coordinate values of 1 to X1a and the Y coordinate values of Y11 to Y1a are stored in the frame line candidate storage unit 17.

With reference to FIG. 10, the contents of the processing for creating a frame corner candidate (step S104) will be described.

In each frame detection area, a vertical line passing through the X coordinate value and a horizontal line passing through the Y coordinate value of the frame line candidate storage unit 17 are drawn, and an image having these vertical and horizontal lines drawn is created. Label the created image. Thereafter, the detected corner coordinates of the closed region are set as corner position candidate coordinates of the character frame and stored in the frame corner candidate coordinate storage unit 18.

FIG. 10A is an explanatory diagram showing an example of an image obtained by drawing a line in each frame detection area obtained in FIG. 9 according to the frame line candidate position. FIG. 10B is an explanatory diagram illustrating the closed regions Lp1 to Lp17 obtained by the labeling. FIG.
The upper left corner coordinate, the lower left corner coordinate, the upper right corner coordinate, and the lower right corner coordinate of the closed regions Lp1 to Lp17 shown in (b) are set as frame corner candidate coordinates, and these are stored in the frame corner candidate coordinate storage unit 18.

Referring to FIG. 11, the contents of the frame position determination process (step S105) will be described.

In the region where the left vertical line frame detection region of the first frame and the upper horizontal line frame detection region of the first frame overlap, the upper left corner candidate coordinates are defined as the first upper left coordinate candidate. I do. Upper left, lower left, upper right and lower right frames closest to the upper left, lower left, upper right and lower right coordinate positions of each character frame relatively determined according to the frame registration data based on each first frame upper left coordinate candidate The distance value from the corner candidate coordinates is accumulated for each of the upper left coordinate candidates of the first frame. The upper left corner, lower left corner, upper right corner, and lower right corner corner candidate coordinates selected when the accumulated value is the minimum are set as the character frame inner coordinates.

FIG. 11A is an explanatory diagram of a case where a character frame is arranged based on the closed area Lp1 in accordance with the character frame registration data. FIG. 11B is an explanatory diagram when a character frame is arranged based on the closed region Lp10 in accordance with the character frame registration data. In the example shown in this embodiment, the closed area Lp1 in the area S shown in FIG. 10B and the upper left corner of the closed area Lp10 are character frame corner candidate coordinates, and based on the closed area Lp1 according to the character frame registration data. FIG. 11A shows a case where a character frame is arranged, and FIG. 11B shows a case where a character frame is arranged on the basis of the closed region Lp10.

The upper left corner of the closed area Lp1 is the upper left corner of the character frame of the form image data. The upper left corner, the lower left, and the upper right of each character frame relatively determined based on the frame registration data based on the upper left corner of the closed area Lp1. And the lower right coordinates are close to other frame corner candidate coordinates and have a smaller distance.

Since the upper left corner of the closed area Lp10 is the frame corner candidate coordinate obtained from the background printing, it has no relation to the frame position, and is relatively determined based on the frame registration data based on the upper left corner of the closed area Lp10. Since no corner candidate coordinates are found at positions close to the upper left, lower left, upper right, and lower right coordinates of each of the character frames, the cumulative value of the distance increases.

As a result, the upper left corner of the closed area LP1 is determined as the frame position reference, and the coordinates of the following positions shown in FIG. 11A are determined as the frame inner positions. Is output to

Upper left coordinate position Upper right coordinate position Lower left coordinate position Lower right coordinate position First frame (X12, Y11) (X12, Y13) (X12, Y13) (X14, Y14) Second frame (X15, Y15) (X17, Y15) (X15, Y17) (X17, Y17) Third frame (X18, Y18) (X1a, Y18) (X18, Y1a) (X1a, Y1a) Then, the character recognition unit 15 Character recognition is performed on the form image data within the detected coordinates inside the frame.

FIG. 12 is an explanatory diagram showing the character image data within the detected coordinates inside the frame, which is the result of the character recognition processing by the character recognition device of FIG. As shown in FIG.
By the above-described character recognition processing, three character frames 19a, 1
The character frame position can be accurately detected from the form image data in a state where the characters 9b and 19c cannot be normally detected, and the character entered in the character frame can be recognized.

As described above, according to the present invention, the frame lines are blurred in the form image data or the background is printed around the character frame depending on the characteristics of the image scanner and the reflectance of the print character frame on the form paper. Also, it is possible to accurately detect the position of the character frame without erroneously determining the background printed around the character frame or the character written in the character frame as the character frame.

Therefore, since the character frame and the character and the background, which cannot be detected by the conventional character frame position detection based on the pixel distribution, can be distinguished, the character frame position can be accurately detected and written in the character frame. Characters can be recognized.

[0055]

As described above, according to the present invention,
When the characters written in the character frame are optically read and recognized, the position of the character frame is detected based on the frame area direction code obtained in the detection area of the character frame. Even if the frame is faint or there is a background print around the character frame, the character frame position can be accurately detected, and the character entered in the character frame can be recognized.

Further, according to the character recognition device of the present invention,
The above character recognition method can be realized.

[Brief description of the drawings]

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

FIG. 2 is an explanatory diagram showing an example of character frame registration data stored in a character frame registration data storage unit.

FIG. 3 is a flowchart illustrating a flow of a character frame position detection process by a character frame position detection unit in FIG. 1;

FIG. 4 is an explanatory diagram showing an example of form image data read by an image scanner unit of FIG. 1 and stored in an image data storage unit;

5A and 5B show a frame detection area by two vertical lines of the form image data, FIG. 5A is an explanatory diagram of a first frame, and FIG.
(C) is an explanatory diagram of the third frame.

6A and 6B show a frame detection area by two horizontal lines of the form image data, FIG. 6A is an explanatory diagram of a first frame, and FIG.
(C) is an explanatory diagram of the third frame.

FIG. 7A is an explanatory diagram showing a frame area direction code when the current trace point is i and the previous trace point is (i-1) in the contour trace, and FIG.
FIG. 3C is an explanatory diagram showing a frame region direction code when the outer contour of the object is traced as an example of an object of 3 pixels, and FIG. 4C is an explanatory diagram showing frame region direction code data stored in a direction code storage unit. is there.

FIG. 8A is an explanatory diagram showing a frame region direction code for a hole region having a size of 2 × 2 in an object, and FIG. 8B is a diagram showing a frame region direction code of FIG. FIG. 4 is an explanatory diagram showing obtained data.

FIG. 9 shows an example of a histogram, and FIG.
FIG. 4 is an explanatory diagram of an example obtained from the first frame, and FIG.
FIG. 3C is an explanatory diagram of an example obtained from the frame, and FIG. 4C is an explanatory diagram of an example obtained from the third frame.

FIG. 10A is an explanatory diagram showing an example of an image in which a line is drawn in each frame detection area obtained in FIG. 9 according to a frame line candidate position,
(B) shows closed regions Lp1 to Lp17 obtained by labeling.
FIG.

11A and 11B show a case where a character frame is arranged according to character frame registration data. FIG. 11A is an explanatory diagram based on a closed region Lp1, and FIG. 11B is an explanatory diagram based on a closed region Lp10. is there.

12 is an explanatory diagram showing the character image data in the detected coordinates inside the frame, which is the result of the character recognition processing by the character recognition device of FIG. 1;

FIG. 13 shows a state in which a character frame cannot be normally detected;
(A) is an explanatory diagram of form image data in which a character frame is blurred, (b) is an explanatory diagram of a histogram obtained by projecting black pixels obtained from the region (a), and (c) is an erroneously detected character. It is explanatory drawing of a frame position.

[Explanation of symbols]

 Reference Signs List 10 Character recognition device 11 Image scanner unit 11 12 Image data storage unit 12 13 Character frame registration data storage unit 13 14 Character frame position detection unit 14 15 Character recognition unit 15 16 Direction code storage unit 16 17 Frame candidate storage unit 17 18 Frame corner Candidate coordinate storage unit 18 19a, 19b, 19c Character frame (X_S, Y_S) Coordinate Dx Distance H Vertical width Lp Closed area S Area W Width

Claims (9)

[Claims] 1. A character recognition method for optically reading a character written in a character frame and recognizing an image, wherein a position of the character frame is detected based on a frame region direction code obtained in a detection region of the character frame. A character recognition method characterized in that: 2. A character frame area determining process for determining a detection area of the character frame from image data obtained by performing image processing on the character frame and character frame registration data including a frame coordinate value. A frame area direction code extraction process for extracting the frame area direction code from the detected detection area, and a frame line candidate detection for obtaining a frame line candidate position that is a candidate for a character frame inside position from the extracted frame area direction code Processing, a frame corner candidate creating process for detecting corner position candidate coordinates of the character frame after performing labeling on an image in which candidate frame lines created based on the frame line candidate positions are written; and From the candidate coordinates and the character frame registration data,
The character recognition method according to claim 1, wherein the detection is performed through a frame position determination process for determining the position inside the character frame. 3. The character according to claim 1, wherein the frame area direction code is obtained by contour tracing an object formed by connecting black pixels in the detection area of the character frame. Recognition method. 4. The character recognition method according to claim 1, wherein said character frame is printed in a non-dropout color. 5. A character recognition device for optically reading a character written in a character frame and recognizing an image, wherein a position of the character frame is detected based on a frame region direction code obtained in a detection region of the character frame. A character recognition device comprising character frame position detection means. 6. The character frame position detecting means determines a detection region of the character frame from image data obtained by performing image processing on the character frame and character frame registration data including a frame coordinate value. The frame region direction code is extracted from the inside, a frame line candidate position that is a candidate for a character frame inside position is obtained from the extracted frame region direction code, and a candidate frame line created based on the frame line candidate position is written. After performing labeling on the image, the corner position candidate coordinates of the character frame are detected, and from the corner position candidate coordinates and the character frame registration data,
The character recognition device according to claim 5, wherein the position inside the character frame is determined. 7. The character according to claim 5, wherein the frame area direction code is obtained by contour tracing an object formed by connecting black pixels in the detection area of the character frame. Recognition device. 8. The character recognition device according to claim 5, wherein a form in which characters are written in a character frame is optically read,
An image scanner unit for obtaining form image data composed of binary data by photoelectric conversion, an image data storage unit for storing form image data output from the image scanner unit, and including a character frame coordinate value registered in advance Character recognition further comprising: a character frame registration data storage unit that stores character frame registration data; and a character recognition unit that performs character recognition based on the image of the character frame output from the character frame position detection unit. apparatus. 9. The character recognition apparatus according to claim 5, wherein said character frame is printed in a non-dropout color.