JP2000222524A - Method and device for removing ruled line, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To remove only a ruled line part while correctly leaving only a character part as it is in the case of overlap of rules lines and characters in a form image. SOLUTION: The form image is inputted in an image input part 1 by an image input means like a scanner, and the coordinates of ruled lines such as solid lines and broken lines in the form inputted in this image input part 1 are specified by a ruled line specifying part 2. Concretely, a long continuity of black pixels in one direction is regarded as a ruled line. If ruled lines specified by the ruled line specifying part 2 come into contact with or cross characters, ruled lines are removed while leaving the character part by a ruled line removing part 3, and the form image after removal of ruled lines is sent to an output. The form image after removal of ruled lines sent from the ruled line removing part 3 is inputted to a character recognition part 4. This character recognition part 4 segments characters from the inputted form image to perform character recognition and word collation and outputs the recognition result and the collation result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character recognition device (Optical character recognition device) for optically reading characters written on paper.
(Character Recognition: OCR device), not related to OCR-only forms printed in colors that cannot be optically identified, but to the recognition technology of the same color characters written on forms made with general black ruled lines. And a technique for removing a ruled line portion while leaving a character stroke in contact with or intersecting the ruled line.

[0002]

2. Description of the Related Art An optical character recognition device (hereinafter, referred to as an "OCR device") for optically reading characters written on a paper is an image input device such as a scanner or a camera. A character portion is cut out from the form image captured by the means, and the cut out character portion is recognized. For example, Japanese Patent Application Laid-Open No. 7-5
The “character extraction method” described in Japanese Patent No. 7047 can be used.

In the OCR apparatus, a dedicated form printed in a color which cannot be optically distinguished is usually used, but a form other than such a dedicated form, for example, a black ruled line used for copying is used. In some cases, it is necessary to recognize characters written on a form or the like. In such a case, if the ruled line and the character stroke partially overlap, the accuracy of character recognition cannot be improved. The "character extraction method" can be applied to a process of removing only ruled lines in such a case. That is, by checking the state of the pixels existing on the outer edge of the circumscribed polygon of the ruled line, the overlap between the character stroke and the ruled line is detected, and only the ruled line is cut out and removed.

However, when an image of a form is actually taken in, a ruled line in the form image may be inclined or may be inclined due to factors such as the layout of the form, mechanical and optical characteristics of the image input means, and quantization error. May become uneven due to noise, distortion or deflection may occur in the ruled line, or the thickness of the ruled line may change in the middle. In these cases, in the above character extraction method, the circumscribed polygon of the ruled line does not match the contour of the ruled line, and the circumscribed polygon is cut out thicker. As a result, the character stroke near the ruled line is also incorrectly cut along with the ruled line. May be removed.

If the circumscribed polygon 12 of the ruled line 11 does not match the contour of the ruled line 11, as in the form image shown in FIG. 12 (a), the circumscribed polygon 12 is cut out thicker. As a result, as shown in FIG. 12B, the character stroke near the ruled line is also removed together with the ruled line 11.
As described above, there is a problem that the inclination of the ruled line cannot be accurately obtained due to the influence of noise or a part of the character stroke overlapping the ruled line.

Further, when checking the black and white state of the pixels existing at the outer edge of the circumscribed polygon of the line segment shown in FIG. 13A, the circumscribed line segment shown in FIG. If the noise 13 exists near the outer edge of the polygon, the noise 13
Is determined as a part of the character, and there is a problem that the image is recognized as an erroneous image as shown in FIG.

Further, as shown in FIG.
In the case where the character stroke intersects at a shallow angle with respect to 1, the outer edge of the circumscribed polygon of the ruled line becomes a white pixel near the center of the intersection as shown in FIG. Therefore, the character portion may be erroneously removed as shown in FIG.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a ruled line removing method capable of effectively removing only pixels in a ruled line portion while reliably retaining pixels constituting a character. Another object of the present invention is to provide a ruled line removing apparatus and a recording medium suitable for implementing the ruled line removing method.

[0009]

According to the present invention, there is provided a ruled line removing method for removing a ruled line from image data in which a ruled line and a character are composed of pixels of the same color while leaving the character. The process of estimating the position of the outline on both sides of the ruled line by detecting the arrangement of the pixel colors, and the process of estimating the position of the outline and the character stroke by checking the color of the pixels outside the estimated outline. The method includes a step of determining the presence or absence of a contact and a step of converting the color of a pixel forming the ruled line to a color representing a blank when there is a character stroke in contact with the ruled line. Through these processes,
Even when the character and the ruled line are in contact, the image data of the character portion can be correctly cut out.

In the process of judging the presence or absence of the contact, for example, the colors of adjacent pixels on both sides of the estimated contour are checked while being shifted one pixel at a time in the line segment direction of the ruled line. Then, on both outer sides of the contour line, regions where pixels adjacent to each other in the line segment direction have the same color (including a case where one is a continuous black and the other is a continuous white) are each determined as a determination region. The presence or absence of contact between the ruled line and the character is determined from the combination of the pixel colors in the determination area. When there is a contact with the character, the form of the contact is specified according to a combination of the pixel colors of the plurality of determination regions on both outer sides of the estimated contour line. In this case, in order to increase the accuracy, in each judgment area,
The pixel color of the determination area is determined by determining a combination condition of a color of a pixel outwardly adjacent to the contour and a color of one or more pixels further outward from the pixel. I do. Alternatively, for each pixel of the line closest to the contour line, the combination condition of the color of the pixel and the color of the pixel that is separated in the outer direction and the color of the pixel that is adjacent in the line segment direction is determined. The pixel color is determined.

In the step of estimating the position of the contour, for example, the pixel area of the ruled line is subdivided in a direction perpendicular to the direction of the line segment, and the statistics of the length of the plurality of subdivided areas in the direction of division are obtained. The thickness of the ruled line is estimated from the value, and the change in the representative coordinate values of the plurality of subdivided regions in which the difference between the length in the dividing direction and the estimated thickness of the ruled line is within a certain threshold value is linearly approximated. Is detected, and the position of the contour of the ruled line is estimated based on the detected inclination.
Alternatively, the contour of the ruled line is estimated by sequentially joining the end contours of a plurality of subdivided regions in which the difference between the length in the dividing direction and the estimated ruled line thickness is within a certain threshold value. Alternatively, the contour of the ruled line is estimated by sequentially joining the end contours of a plurality of subdivided regions having the same length in the dividing direction and the estimated ruled line thickness.

According to another aspect of the present invention, there is provided a ruled line removing apparatus for inputting image data in which a ruled line and a character are constituted by pixels of the same color; Ruled line identifying means for identifying the ruled line and its position by detecting the arrangement of the pixel colors to be detected, and estimating the contour of the identified ruled line and checking the color of the pixels outside the contoured line to determine the ruled line. Ruled line removing means for determining the presence / absence of contact with a character, and converting a pixel constituting the ruled line into a pixel representing a space when there is a character in contact with the ruled line; Can be cut out.

A recording medium according to the present invention for solving the above-mentioned other problems is a computer-readable recording medium in which a program code for causing a computer device to execute the following processing is recorded. (1) processing of inputting image data in which ruled lines and characters are composed of pixels of the same color; (2) specifying ruled lines and their positions by detecting an array of pixel colors included in the input image data (3) Estimating the contour of the specified ruled line and checking the color of the pixels outside the contoured line to determine whether or not the ruled line has touched a character, and (4) the ruled line When there is a character that is in contact with, a process of converting pixels constituting the ruled line into a color representing a blank.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is applied to an OCR device will be described below with reference to the drawings. FIG. 1 is a functional configuration diagram of the OCR device according to the present embodiment. The OCR device is realized by, for example, an image capturing device such as a scanner, a display device, a data input device such as a keyboard, and a computer device having an external storage device. More specifically, the computer includes functional blocks of an image input unit 1, a ruled line specifying unit 2, a ruled line removing unit 3, and a character recognizing unit 4 formed by reading and executing a predetermined program code. Be composed. The portion of the OCR device except for the character recognition unit 4 functions as a ruled line removing device.

The program code is normally stored in a computer device or in an external storage device, and is read and executed as needed. However, if the functional blocks 1 to 4 are formed, the present invention is implemented. Since it can be implemented, the recording form may be arbitrary. For example,
CD-ROM, DV separable from the computer device
D, FD, or other portable recording media, or recorded on a program server or the like connected to the local network, read at the time of use, installed in the computer device or in an external storage device, and provided for execution at any time It may be.

The function blocks 1 to 4 may be formed by the program code alone, or may be formed appropriately in cooperation with an operating system or another application program mounted on a computer device. Good.

The function blocks 1 to 4 execute the following functions. The image input unit 1 inputs image (image) data captured by the image capturing device for subsequent processing. In the present embodiment, as an example of image data, a form image in which characters are written in a form created with a general black ruled line is used.

The ruled line specifying unit 2 specifies ruled lines such as solid lines and broken lines in the input form image and their positions (coordinate values). As a method of specifying the coordinate values in this case, for example, a method in which a black pixel that is long and continuous in one direction is set as a ruled line, and the starting and ending points of the ruled line and the coordinate values of the contour thereof are specified. A method of projecting in the vertical direction and using the coordinate value of the peak position of the projected histogram as the coordinate value of the ruled line can be used. All of these techniques are known.

The ruled line removing unit 3 determines whether the ruled line specified by the ruled line specifying unit 2 has a contact or an intersection of character strokes (hereinafter, referred to as “contact” unless it is necessary to distinguish the two). Is detected, and if there is a contact, only the ruled line portion is removed while leaving the character stroke.
The details of the ruled line removal processing will be described later.

The character recognition section 4 cuts out a character portion from the form image from which the ruled line has been removed by the ruled line removing section 3 and performs character recognition and word collation of the cut out portion.

Next, detailed functional blocks of the ruled line removing section 3 will be described with reference to FIG. In the ruled line removing unit 3,
The form image and the coordinate values of the ruled lines extracted by the ruled line specifying unit 2 are input to the ruled line contour estimating unit 31.
The ruled line contour estimating unit 31 estimates the contours on both sides of the ruled line, and causes the character contact / intersection determining unit 32 to input the estimation result. The character contact / intersection determination unit 32 checks the color of the pixel outside the outline, whether it is white or black in the case of a monochrome image, based on the estimated outline, and determines the contact between the ruled line and the character stroke. Determine the presence or absence. The removal image creation unit 33 creates a ruled line removal image by converting the black pixels in the ruled line portion into white pixels while leaving the character portion in accordance with the determination result by the character contact / intersection determination unit 32.

Here, as an example, a specific processing content of the ruled line removing unit 3 for removing a horizontal ruled line from a form image, that is, a ruled line removing method will be specifically described. In the case of a vertical ruled line, the same processing can be performed by replacing the coordinate values of the form image from horizontal to vertical.

<Judgment of Contact Existence> FIG. 3 is an explanatory diagram of a method of judging the presence or absence of character contact in the character contact / intersection judgment unit 32. Here, pixels on the upper and lower sides of the contour 35 of the ruled line 34 estimated by the ruled line contour estimating unit 31 are determined as decision points 36a and 36b, and the decision points 36a and 36b are determined.
While shifting b one pixel at a time in the ruled line direction (right direction in the figure), it is checked whether the judgment points 36a and 36b are white pixels or black pixels. In addition, the upper and lower decision points 36a, 3
Assuming that areas including the same combination between the right and left adjacent judgment points are the judgment areas 37, the judgment points 36a, 36a,
There are four combinations of 36b: white / white, white / black, black / white, and black / black.

Referring to the other judgment areas adjacent to the left and right of the judgment area 37 of interest and looking at their combinations, they are classified as shown in FIG. 4, for example. 4A is a region where the character stroke does not touch the ruled line, FIG. 4B is a region where the character stroke is in contact with the ruled line, and FIG. FIG. 4D shows a region where a character stroke crosses a ruled line and a character stroke is present only on one of the upper and lower sides of the ruled line.

<Removal of Ruled Line> The procedure for removing the ruled line in each area shown in FIG. 4 is as follows. In the drawing, in the determination area 37, the area indicated by halftone is a black or white area, the area indicated by white is an area only white, and the area indicated by black is an area only black. A hatched portion of the ruled line 34 is a portion before the pixel is removed, a blank portion is a portion from which the pixel is removed, and a black portion is a portion of a pixel to be left.

In the case of FIG. 4A: all the black pixel portions of the ruled line 34 corresponding to the judgment area 37 are converted into white pixels. In the case of FIG. 4B, as the touched character portion, the length of the base is equal to the length of the black pixel portion of the determination area 37, and the height is equal to half the thickness of the ruled line 34. The black pixels of the ruled line 34 are converted to white pixels so that the triangle remains. However, this process is only an example, and other methods such as removing the entirety of the inside of the determination area 37, removing only half the width of the ruled line 34, or leaving all the other way around can be used. . In the case of FIG. 4C, the character part intersects the ruled line 34, and the black pixels of the ruled line 34 in the area where the character part is seen both above and below the ruled line 34 are left as they are. In FIG. 4D, the black pixels of the ruled lines 34 of the intersecting character portions are left as they are, and the black pixels are converted into white pixels only in the portions where the character portions do not intersect.

<Estimation of outline when ruled line is inclined> When the outline of the ruled line is estimated when the ruled line is inclined, first, the thickness of the ruled line is estimated. When estimating the thickness of a ruled line, for example, part or all of the ruled line is divided into two connected components in which black pixels are connected in a direction perpendicular to the ruled line. The divided two connected components form a set of black pixels having the same coordinate value in the ruled line direction. Statistics of the lengths of these two connected components are obtained, and the average value is estimated as the ruled line thickness. Note that a mode, a median, or the like may be used instead of the average.

When the thickness of the ruled line is known, the inclination of the ruled line is detected as follows. First, as shown in FIG. 5, part or all of the ruled line is divided into two connected components in which black pixels are connected in a direction perpendicular to the ruled line. Then, of the divided two connected components, only those whose length of the two connected components matches the estimated rule line thickness are extracted. Thereafter, the center coordinate values of the extracted two connected components are linearly approximated by, for example, the least squares method, and the inclination thereof is obtained. Instead of the center coordinate value, a representative coordinate value uniquely defining a specific position of the two connected components, for example, a coordinate value of one of the end points may be used. In FIG. 5, a portion indicated by halftone is a two-connected component having the same length as the estimated ruled line thickness, and a portion indicated by white is a two-component whose estimated ruled line thickness and length are different.
It is a connected component.

The outline of the inclined ruled line is as shown in FIG.
Estimation is performed based on the circumscribed polygon of the ruled line, the coordinate value of the ruled line obtained by the ruled line specifying unit 2, the estimated thickness and inclination of the ruled line.

FIG. 7 is a diagram showing an outline of processing for more accurately estimating the outline of a ruled line based on the estimated thickness of the ruled line. Here, a part or all of the ruled line is divided into two connected components in which black pixels are connected to the ruled line in the vertical direction, and the length of the two connected components in the divided two connected components is the estimated ruled line. Only those that match the thickness of are extracted. Then, a line connecting the end points of all the extracted two connected components in sequence is estimated as the outline of the ruled line. Note that, in FIG. 7, the portions indicated by halftones and the portions indicated by white have the same meaning as in FIG.

<Case where Rule Line Thickness Changes> FIG. 8
(A), (b) is a figure for showing the corresponding processing when the thickness of the ruled line 34 changes in the middle. In this case, the thickness of the ruled line 34 is dynamically determined, and the contour is estimated. Specifically, part or all of the ruled line 34 is divided into two connected components in which black pixels are connected to the ruled line 34 in the vertical direction. As shown in FIG. 8A, when attention is paid to a certain two connected components, the ruled lines 34 within a certain range centered on the two connected components.
Estimate the thickness of. Next, the estimated thickness of the ruled line 34 and
Two connected components having the same length are extracted. As shown in FIG. 8B, while moving the two connected components of interest one pixel at a time in the ruled line direction, it is determined whether to extract or not to extract all the two connected components constituting the ruled line 34. Then, the outline of the ruled line 34 is estimated by sequentially connecting the end points of all the extracted two connected components.

<Case where noise is mixed> Even when the contour of the ruled line is uneven due to the mixed noise, in order to accurately determine the contact portion of the character, a plurality of lines outside the estimated contour are determined. The contact or intersection between the ruled line and the character portion is determined with reference to the pixel. The processing in this case will be described with reference to FIG. 9 and FIG. 10 taking the case of a horizontal ruled line as an example.

FIG. 9 and FIG. 10 show examples in which two lines of pixels are referred to for each of the judgment points above and below the ruled line. The one closer to the ruled line is an inner judgment point, and the one farther from the ruled line is an outer judgment point. In the example of FIG. 9, when both the inner determination point and the outer determination point are black, the determination point is determined to be black. On the other hand, when one or both of the pixels at the inner judgment point and the outer judgment point are white, the judgment point is regarded as white. Thus, in FIG.
FIG. 3 shows a case where a determination is made using one outer determination point. On the other hand, in the case of FIG. 10, when attention is paid to each coordinate value in the ruled line direction, for one inside judgment point, three outside judgment points, the coordinate value of interest and the coordinate values before and after it, are regarded as outside judgment points. .

For example, the inner judgment point is black and 3
When at least one of the outer determination points is black, the determination point of the target coordinate value is set to black. When the inner judgment point is white or when all three outer judgment points are white, the target judgment point is judged to be white. In this way, the black and white of the determination point is determined from a plurality of lines outside the estimated contour, and the presence or absence of the contact or intersection of the character with the ruled line is determined in the same manner as the above-described method of determining the contact or intersection of the character portion. 9 and 10 are examples based on the combination of the colors of the pixels on two lines.
The pixel color of the determination point may be determined according to a combination of the colors of the pixels on three or more lines. In this case, as the criterion of the pixel color, the above-described logical condition may be used, or a majority decision logic or other combination condition may be used. In this way, it is possible to further improve the accuracy of determining whether a character touches or intersects a ruled line.

<Case where character stroke intersects ruled line at a shallow angle> Considering the length of continuous white state on both sides of the ruled line, character strokes intersect the ruled line at a shallow angle. Even if there is, contact or intersection of the character portion with the ruled line can be accurately determined. Hereinafter, the processing content in this case will be described by taking a case of a horizontal ruled line as an example.

The black-and-white state (upper / lower) of a certain judgment area 37 is white / black or black / white, and the black-and-white state (upper / lower) of at least one of the judgment areas 37 is white / white. If the length is less than a certain threshold, it is determined that the character stroke intersects the ruled line shallowly. FIGS. 11A to 11J show examples of this case. The ruled line removal method is described by referring to a determination area next to a determination area whose white / white state (upper / lower) is white / white and whose length is equal to or less than a threshold value, for example, as shown in FIG. Then, the ruled line is removed while leaving the character portion. In FIG. 11, in the judgment area, the black state indicates black, the white state indicates white length greater than or equal to the threshold, the halftone state indicates black or white, and the white circle indicates white length less than the threshold. In the ruled line portion, the halftone state indicates a state before removal, the blank state indicates a removed part, and the black state indicates a part where a ruled line is left.

As described above, according to the OCR apparatus of the present embodiment, when the ruled line is inclined, when the contour of the ruled line is uneven due to noise, when the ruled line is distorted or bent, or Even if the thickness of the ruled line changes in the middle, the presence or absence of contact between the ruled line and the character stroke can be determined. Further, even when the angle of the character stroke that intersects the ruled line is small, the intersection between the ruled line and the character stroke can be accurately determined. This allows
The accuracy of character segmentation is increased, and the recognition accuracy of the character recognition unit 4 can be significantly improved as compared with the conventional method.

In this embodiment, for convenience of explanation, an example is described in which image data composed of a combination of white pixels and black pixels is used. In addition, the present invention can be applied to all image data composed of a combination of pixel colors capable of expressing a blank color.

[0039]

As is apparent from the above description, according to the present invention, a unique effect is obtained in that only the ruled line portion can be effectively removed while the character portion is securely left.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a block diagram showing details of a ruled line removing unit.

FIG. 3 is an explanatory diagram showing a procedure for determining whether a character touches or intersects;

4A is an area where a character does not touch or cross a ruled line, FIG. 4B is an area where a character is touching a ruled line, and FIG.
Is an area where characters intersect the ruled line and there is a character portion both above and below the ruled line, and (d) is an image of a form before and after removal in an area where a character intersects the ruled line and there is only a character portion above and below the ruled line. FIG.

FIG. 5 is an explanatory diagram showing a procedure for detecting the inclination of a ruled line.

FIG. 6 is an explanatory diagram showing a point of estimating a contour of an inclined ruled line.

FIG. 7 is an explanatory diagram showing a point of estimating a contour of a ruled line;

FIGS. 8A and 8B are explanatory diagrams showing a procedure for dynamically detecting the thickness of a ruled line when the thickness of the ruled line changes on the way; FIGS.

FIG. 9 is an explanatory diagram showing a procedure for determining whether a ruled line touches or intersects a character with reference to pixels on a plurality of lines outside the contour;

FIG. 10 is an explanatory diagram showing a procedure for determining whether a ruled line touches or intersects with a character by referring to pixels on a plurality of lines outside a contour;

FIGS. 11A to 11J are explanatory diagrams showing states before and after ruled line removal when a character stroke intersects a ruled line shallowly;

FIGS. 12A and 12B are diagrams illustrating a state in which the inclination of a ruled line cannot be accurately obtained when there is noise or a character overlapping the ruled line;

13A is a ruled line containing noise, FIG. 13B is a partially enlarged view of the ruled line, and FIG. 13C is an explanatory diagram showing a state where a noise component is erroneously recognized as a character.

14A is a diagram showing a state in which a character stroke intersects a ruled line at a shallow angle, FIG. 14B is a partially enlarged view thereof, and FIG. 14C is a diagram in which a character portion is erroneously removed in this case. FIG.

[Explanation of symbols]

 1 image input unit 2 ruled line specifying unit 3 ruled line removing unit 4 character recognizing unit 31 ruled line contour estimating unit 32 character contact / intersection determining unit 33 removed image creating unit

Claims (10)

[Claims] 1. A method for removing a ruled line while leaving the character from image data in which the ruled line and the character are composed of pixels of the same color, wherein the contour of both sides of the ruled line is detected by detecting an array of pixel colors. Estimating the position of the line, determining the presence or absence of contact between the ruled line and the character stroke by checking the color of the pixel outside the estimated contour line, and determining the character stroke in contact with the ruled line Converting the color of the pixels forming the ruled line to a color representing a blank when there is a ruled line. 2. The step of judging the presence / absence of contact includes checking the color of adjacent pixels on both sides of the estimated contour line while shifting the color by one pixel in the direction of the line segment of the ruled line.
Areas in which pixels adjacent to each other in the line segment direction on both outer sides of the contour line have the same color are determined areas, respectively.
2. The ruled line removal method according to claim 1, wherein the presence / absence of contact between the ruled line and the character is determined based on a combination of pixel colors in each determination region. 3. When there is a contact of the character, a contact mode is specified according to a combination of pixel colors of a plurality of the determination areas on both outer sides of the estimated contour line. The ruled line removing method according to claim 3. 4. In each determination area, a combination condition of a color of a pixel outwardly adjacent to the contour line and a color of a pixel further outward by one or more pixels from the pixel is determined. 4. The ruled line removing method according to claim 2, wherein a pixel color of the determination area is determined by performing the determination. 5. For each pixel of a line closest to the contour line, a combination condition of a color of the pixel and a color of a pixel which is separated in an outward direction and a pixel which is adjacent in a line segment direction is determined. 5. The ruled line removal method according to claim 4, wherein a pixel color of the determination area is determined. 6. The step of estimating the position of the contour line includes subdividing a pixel region of the ruled line in a direction perpendicular to a line segment direction and calculating a length of a plurality of subdivided regions in a dividing direction. The thickness of the ruled line is estimated from the value, and the change in the representative coordinate values of the plurality of subdivided regions in which the difference between the length in the dividing direction and the estimated thickness of the ruled line is within a certain threshold value is linearly approximated. The ruled line removing method according to any one of claims 1 to 5, wherein a slope of the ruled line is detected, and a position of an outline of the ruled line is estimated based on the detected slope. 7. The step of estimating the position of the contour line includes subdividing a pixel region of the ruled line in a direction perpendicular to a line segment direction and calculating a length of a plurality of subdivided regions in a dividing direction. The thickness of the ruled line is estimated from the value, and the end contours of a plurality of subdivided regions in which the difference between the length of the ruled line and the estimated ruled line thickness is within a certain threshold value are sequentially joined to form the ruled line. The ruled line removal method according to any one of claims 1 to 5, wherein the contour is estimated. 8. The step of estimating the position of the contour line includes subdividing a pixel region of the ruled line in a direction perpendicular to a line segment direction and calculating a length of a plurality of subdivided regions in a dividing direction. Estimating the thickness of the ruled line from the values, and estimating the outline of the ruled line by sequentially joining the end contours of a plurality of subdivided regions in which the length in the dividing direction is equal to the estimated ruled line thickness. The ruled line removal method according to any one of claims 1 to 5, characterized in that: 9. An image data input means for inputting image data in which a ruled line and a character are composed of pixels of the same color, and a ruled line and its image data are detected by detecting an array of pixel colors contained in the input image data. Ruled line specifying means for specifying a position; estimating a contour of the specified ruled line; checking colors of pixels outside the contoured line to determine whether or not the ruled line is in contact with a character; A ruled line removing device for converting a pixel forming the ruled line into a pixel representing a blank when there is a touching character, and wherein only the pixel data of the character portion can be cut out; 10. A process of inputting image data in which a ruled line and a character are composed of pixels of the same color, and specifying a ruled line and its position by detecting an array of pixel colors included in the input image data. Processing, estimating the outline of the identified ruled line and checking the color of pixels outside the outline to determine whether or not the ruled line is in contact with the character; A computer-readable recording medium in which a program code for causing a computer device to execute a process of converting a pixel constituting the ruled line into a pixel representing a blank at a certain time is recorded.