JP2008147976A - Image inclination correction device and image inclination correcting method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image inclination correction device capable of performing high-speed processing and correctly detecting outer edges of a slip even when there is a partial omission or a painted-out area in an outer edge part of the slip. <P>SOLUTION: The image inclination correction device to be disclosed comprises: an edge position generating part 203 for generating the position of an edge point of outer edges of a slip from data of a line buffer 202; an edge position storage memory 210 for storing information of the edge position; an edge candidate detecting part 205 for using the stored information of the edge position to detect a plurality of edge straight line candidates to be the outer edges of the slip; an edge selecting part 206 for selecting a combination of four appropriate sides as a slip from the edge straight line candidates; a corner detecting part 207 for detecting corner coordinates of the slip; a parameter generating part 208 for generating parameters needed when correcting a slip image from the corner coordinates; an image storage memory 204 for storing information of the line buffer as an image; and an image correcting part 209 for using the parameters to correct the image of the image storage memory. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image inclination correction apparatus and an image inclination correction method for correcting the inclination of a read form and correctly cutting out a form area in a scanner apparatus or the like having a paper feed mechanism.

  In a scanner apparatus having a paper feed mechanism, output data is raster image data, and an area outside the form image is also read at the same time. Therefore, when the form to be read is not correctly arranged with respect to the scanner apparatus The outer edge of the output form image is also inclined with respect to the outer edge of the raster image data. Further, if the paper feeding of the scanner device is not performed correctly, the outer edge of the output form image may be inclined with respect to the outer edge of the raster image data even when the form to be read is correctly arranged.

Therefore, in order to generate an image in which unnecessary portions outside the form image are removed, it is necessary to correct the inclination of the form image and cut out only the form area. It is necessary to accurately detect the coordinates of the four corners.
Conventionally, in order to detect the coordinates of the four corners of a form, the form image is temporarily stored in an image memory, and the stored form image is searched in the main scanning direction and the sub-scanning direction to obtain the form image. By detecting a plurality of boundary points (edge points) between the form image and its outside on the upper side, the lower side, the left side, and the right side, and obtaining the intersection of four sides each consisting of a straight line defined by the plurality of edge points, A method for obtaining corner coordinates is known.

However, the output data of the scanner device is a raster image, and is obtained from the scanner device by performing sub-scanning that sequentially moves the scanning position in the vertical direction while repeating scanning in the main scanning direction for the form image. Data is sent. Therefore, in the above detection method, in order to detect the edge points of the upper side and the lower side, the form image must be temporarily stored in the image memory, and a plurality of edge points must be detected by searching in the sub-scanning direction. However, since the search in the sub-scanning direction takes more time than the search in the main scanning direction, there is a problem that the time required for detecting the corner coordinates of the form image becomes long.
In addition, when the corner of the form is bent or cut obliquely or roundly, or as shown in FIG. 17, there is a solid area where the difference in brightness from the outside of the form image is small. In such a case, there is a problem that the form area cannot be detected correctly.

In order to solve such a problem, it is conceivable to correct the inclination of the form image by the following procedure.
A form image input as a raster image is divided into a plurality of ranges at equal intervals in the main scanning direction, and the upper and lower sides of the form are obtained by using the number of pixels (white pixels) representing the form portion existing in each divided range. The position of the edge point (edge position) is detected, and the form image is changed according to the change point at which the pixel representing the outer portion of the form (black pixel) changes to a white pixel and the change point at which the white pixel changes to a black pixel. The edge positions of the left side and right side of are detected. Then, a plurality of edge straight line candidates are obtained as straight lines passing through a plurality of edge points on the same side, and a combination of four sides of the upper side, the lower side, the left side, and the right side is generated using the inclination information of the edge straight line candidates. Then, select the combination of the four sides that maximizes the sum of the distance between the upper side and the lower side and the distance between the left side and the right side, and calculate the corner coordinates of the form by finding the intersection of the straight lines of the selected four sides. To detect. Then, parameters necessary for correcting the image are obtained from the corner coordinates of the form, and the inclination of the form image and the cutout of the form area are performed using these parameters.

On the other hand, in Patent Document 1, the coordinate values of the left edge point and the right edge point of the form image are specified using the main scanning direction as a processing unit, and this processing is repeated in the sub scanning direction for all lines. An interval between the left and right edge points is obtained based on the coordinate value specified by the execution, and a region where the interval is constant in the sub-scanning direction is set as a stable region. A technique for deriving the inclination angle of the form image with respect to the reference position based on the coordinate values of the left edge point and the right edge point existing in the region other than the stable region is disclosed.
However, in Patent Document 1, a form image is divided into a plurality of ranges at equal intervals, and the edge positions of the upper and lower sides of the form are detected by using the number of white pixels representing the form portion existing in each divided range. Then, a plurality of edge straight line candidates are obtained using the detected edge positions, and a combination of four sides of the upper side, the lower side, the left side, and the right side is generated using the inclination of the edge straight line candidate. Select the combination of the four sides that maximizes the sum of the distance between the lower side and the distance between the left side and the right side, find the intersection of the lines using the selected four straight lines, detect the corner coordinates of the form, Thus, there is no description about obtaining parameters necessary for image correction and correcting the form image using these parameters.

Further, in Patent Document 2, the image data of the paper input through the image input device is developed in the image memory in units of pixels, and black and white inversion is performed in the image memory by scanning from at least one direction in the vertical direction or the horizontal direction. Detect the coordinates that are being used. Then, the first black / white reversal coordinate is used as a start position, and a straight line in which the black / white reversal coordinate continues from that position in one direction is detected as one side of the outline of the paper, and the inclination angle of the straight line is used as the paper inclination angle A technique for calculating is disclosed.
However, in Patent Document 2, the form image is divided into a plurality of ranges at equal intervals, and the edge positions of the upper and lower sides of the form are detected using the number of white pixels representing the form portion existing in each divided range. Then, by using the slope of the edge straight line candidate, a combination of the upper side, the lower side, the left side, and the right side is generated, and the sum of the distance between the upper side and the lower side and the distance between the left side and the right side is maximized. There is no description about selecting a combination of four sides and correcting the form image using the parameters obtained from the combination.

In Patent Document 3, the white run length of each scanning line of an image is measured, the measured white run lengths are compared within a predetermined plurality of scanning lines, and the minimum white run length and the scanning line position are determined. Extracts the edge of the original by extracting the straight line that can best connect a plurality of image edge points as the image edge line and calculating the inclination of the image edge line from the vertical line. Technology is disclosed.
However, in Patent Document 3, a combination of four sides of an upper side, a lower side, a left side, and a right side is generated using the slope of the edge straight line candidate, and the distance between the upper side and the lower side, and the distance between the left side and the right side from among them. There is no description about detecting a corner coordinate of a form by selecting a combination of four sides that maximizes the sum of and calculating the intersection of the straight lines using the selected straight lines of the four sides.
Japanese Patent Application Laid-Open No. 07-334618 (FIGS. 1 (a), (b), [0005] to [0009]) Japanese Patent Laying-Open No. 2005-165822 (FIG. 2, [0026], [0027]) JP 02-15882 (FIG. 1, page 580, upper left column, line 19 to upper right column, line 19) Makoto Nagao: Image recognition theory, p72-74, Corona (Showa 58)

  In the conventional image tilt correction apparatus and image tilt correction method, the raster image output from the scanner device is temporarily stored in the image memory, and the search in the main scanning direction and the sub-scan that takes more time than the main scanning direction are processed. Since the direction of the image is detected by performing the direction search, the time required for detecting the inclination of the form image becomes longer, and the number of edge points on the upper and lower sides of the form that can be detected becomes smaller as the inclination of the form image becomes smaller. Therefore, sufficient edge points for detecting the upper side and the lower side cannot be detected stably. In addition, when detecting the upper and lower edges, the accuracy of detecting the stable area is important, but if the form corner is bent or cut diagonally or rounded, there is little difference in brightness at some of the edges of the form. When there is a solid area, a stable area cannot be detected correctly.

  The present invention has been made in view of the above circumstances, and in a scanner apparatus having a paper feed mechanism, a search in the sub-scanning direction is not required, and processing is performed simultaneously with the input of a raster image in the main scanning direction as a processing unit. Therefore, even when the inclination of the form image is small, the corner coordinates of the form image can be obtained accurately, and the inclination of the form image can be corrected. Detection is not required, and when the form corner is bent or cut diagonally or rounded, there is a solid area with little difference in brightness between the outside of the form image and part of the edge of the form An object of the present invention is to provide an image tilt correction apparatus and an image tilt correction method capable of correctly detecting the outer edge of a form image even in some cases.

  In order to solve the above-mentioned problem, an invention according to claim 1 relates to an image tilt correction apparatus, and uses a form read from a line buffer that temporarily stores an inputted form image for each scanning line, and uses the form. An edge position generation unit that generates the positions of the edge points of the upper side and the lower side and the positions of the edge points of the left side and the right side of the form, an edge position storage memory that sequentially stores the information of the generated edge positions, Using the edge position information stored in the edge position storage memory, an edge candidate detection unit that detects a plurality of straight lines as candidates for the edge straight lines on the four sides of the form, and the detected plurality of straight lines. An edge selection unit that selects a combination of four sides that is appropriate as a form from the combination of four sides, and a code that detects the corner coordinates of the form from the combination of the selected four sides. A detection unit, a parameter generation unit that generates parameters necessary for correcting the form image using the detected corner coordinates, an image storage memory that sequentially stores information of the line buffer and stores the information as an image, And an image correction unit that corrects an image stored in the image storage memory using the generated parameter.

  According to a second aspect of the present invention, there is provided an image tilt correction apparatus, wherein an image read from a line buffer that temporarily stores an input form image for each scanning line is used to detect the upper and lower sides of the form. An edge position generation unit that generates positions of edge points and positions of edge points on the left and right sides of the form; an edge position storage memory that sequentially stores information on the generated edge positions; and the edge position storage memory Using the stored edge position information, an upper edge candidate detection unit that detects a plurality of straight lines as edge straight line candidates forming the upper side of the form; and edge position information stored in the edge position storage memory; Detecting a plurality of straight lines as edge straight line candidates that form the lower side, left side, and right side of the form from the information on the upper side edge straight line detected by the upper side edge candidate detection unit. , Left side, right side edge candidate detection unit, an edge selection unit that selects a combination of four sides that is valid as a form from a combination of four sides generated from the detected plurality of straight lines, and the selected four sides A corner detection unit that detects corner coordinates of the form from the combination, a parameter generation unit that generates parameters necessary for correcting the form image using the detected corner coordinates, and information of the line buffer are sequentially stored. An image storage memory for storing an image and an image correction unit for correcting the image stored in the image storage memory using the generated parameters are provided.

  The invention according to claim 3 relates to the image tilt correction apparatus according to claim 1 or 2, wherein the edge position generation unit divides the image data stored in the line buffer into a plurality of ranges at equal intervals. When it is first detected that the number of pixels of the form part existing in each range is 1/2 or more of the number of pixels in the range, the image data is set as the position of the upper edge point at the midpoint of the range. When the number of pixels in the form part existing in each range is finally detected to be 1/2 or more of the number of pixels in the range, The upper and lower edge position generation unit that performs the process of setting the image data number for each line with the point as the position of the lower edge point, and the image data are searched in the line direction, and the pixels in the form part are continuous beyond the form width. Depending on the first pixel position It consists of a left side / right side edge position generation unit that detects the position of the side edge point and detects the position of the right side edge point based on the last pixel position for each line. The edge position storage memory is configured to store the edge position storage memory.

  According to a fourth aspect of the invention, there is provided the image tilt correction apparatus according to the first or third aspect, wherein the edge candidate detecting unit performs Hough conversion from the positions of the four edge positions stored in the edge position storage memory. Is used to detect a plurality of straight lines that are candidates for edge straight lines on four sides.

  According to a fifth aspect of the present invention, in the image tilt correction apparatus according to the second or third aspect, the upper edge candidate detection unit performs Hough conversion from information on the edge position of the upper edge stored in the edge position storage memory. Is used to detect a plurality of straight lines that are candidates for edge straight lines on the upper side.

  According to a sixth aspect of the invention, there is provided the image inclination correcting apparatus according to the fifth aspect, wherein the lower side, left side, and right side edge candidate detection unit is information on edge positions of four sides stored in the edge position storage memory. From the plurality of straight lines that are candidates for edge straight lines of four sides detected by using the Hough transform and the plurality of straight lines that are candidates for edge straight lines detected by the upper edge candidate detection unit, the lower side and the left side , A plurality of straight lines that are candidates for edge straight lines on the right side are detected.

  The invention according to claim 7 relates to the image inclination correction apparatus according to any one of claims 1 to 6, wherein the edge selection unit determines all combinations of four sides having the same angle between two adjacent sides. Among the combinations of all detected four sides, the combination of the four sides that maximizes the sum of the distance between the upper side and the lower side and the distance between the left side and the right side is regarded as a valid four-side combination as the above-mentioned form. It is characterized by being configured to select.

  The invention according to claim 8 relates to the image tilt correction apparatus according to any one of claims 1 to 7, wherein the corner detection unit has four sides selected as an appropriate combination of four sides as the form. It is characterized in that the corner coordinates of the form are detected as the intersection of straight lines (including the case where the straight line is an extension) of two adjacent sides in the edge straight line.

  The invention according to claim 9 relates to the image tilt correction apparatus according to any one of claims 1 to 8, wherein the parameter generation unit moves the coordinates of the input image to a specific position, and at a specific magnification. After the image is enlarged or reduced and rotated by a specific angle, the above parameters are generated by an inverse matrix of a transformation matrix for obtaining an output image by performing an operation for aligning the rotation center with the origin of the coordinates of the output image. It is characterized by that.

  The invention according to claim 10 relates to the image tilt correction apparatus according to any one of claims 1 to 9, wherein the image correction unit corrects the tilt of the input form image using the parameter, The present invention is characterized in that the correction is made so as to remove the outer portion of the form image.

  According to an eleventh aspect of the present invention, there is provided an image tilt correction method, wherein an edge position generation unit uses an image read from a line buffer that temporarily stores an input form image for each scanning line, The position of the edge points on the upper and lower sides of the form and the positions of the edge points on the left and right sides of the form are generated, and the generated edge position information is sequentially stored in the edge position storage memory, and the edge candidate detection means Then, using the edge position information stored in the edge position storage memory, a plurality of straight lines are detected as candidates for the edge straight lines on the four sides of the form, and the edge selecting means detects from the detected plurality of straight lines. A combination of four sides that is appropriate as a form is selected from the generated four side combinations, and the corner detection means selects the corner coordinates of the form from the selected combination of the four sides. And the parameter generation means generates parameters necessary for correcting the form image using the detected corner coordinates, and the image correction means uses the generated parameters to obtain the information of the line buffer. It is characterized by correcting the images stored in the image storage memory that are sequentially stored and stored as images.

  The invention according to claim 12 relates to an image inclination correction method, wherein an edge position generation unit uses an image read from a line buffer that temporarily stores an input form image for each scanning line, The upper edge and lower edge edge point positions and the left and right edge point positions of the form are generated, the generated edge position information is sequentially stored in the edge position storage memory, and the upper edge candidate detection means However, using the edge position information stored in the edge position storage memory, a plurality of straight lines are detected as edge straight line candidates forming the upper side of the form, and the lower side, left side, and right side edge candidate detecting means are From the edge position information stored in the edge position storage memory and the upper edge straight line information detected by the upper edge candidate detection means, the lower side, left side, and right side of the form A plurality of straight lines are detected as edge straight line candidates to be formed, and the edge selecting unit selects a combination of four sides that are valid as a form from the combination of four sides generated from the detected plurality of straight lines, and a corner detecting unit However, the corner coordinates of the form are detected from the selected combination of the four sides, the parameter generation means generates parameters necessary for correcting the form image using the detected corner coordinates, and the image correction means Using the generated parameters, the image in the image storage memory in which the information in the line buffer is sequentially stored and stored as an image is corrected.

  The invention according to claim 13 relates to the image inclination correction method according to claim 11 or 12, wherein, in the edge position generation means, the upper side / lower side edge position generation means stores the image data stored in the line buffer. When it is first detected that the number of pixels of the form part existing in each range is equal to or more than ½ of the number of pixels in the range divided into a plurality of ranges at equal intervals Lastly, the number of pixels in the form part existing in each range is equal to or more than ½ of the number of pixels in the range. When detected, the image data number is set for each line using the middle point of the range as the position of the lower edge point, and the left / right edge position generation means searches the image data in the line direction to generate a form. portion Four edge positions generated by detecting the position of the left edge point by the first pixel position where the pixels are continuous over the form width and detecting the position of the right edge point by the last pixel position for each line. This information is stored in the edge position storage memory.

  The invention according to claim 14 relates to the image tilt correction method according to claim 11 or 13, wherein the edge candidate detection means performs Hough conversion from the information on the edge positions of the four sides stored in the edge position storage memory. Is used to detect a plurality of straight lines that are candidates for edge straight lines on four sides.

  The invention according to claim 15 relates to the image inclination correction method according to claim 12 or 13, wherein the upper edge candidate detection means performs Hough conversion from information on the edge position of the upper edge stored in the edge position storage memory. Is used to detect a plurality of straight lines that are candidates for the upper edge straight line.

  The invention according to claim 16 relates to the image inclination correction method according to claim 15, wherein the lower side, left side, and right side edge candidate detecting means is information on edge positions of four sides stored in the edge position storage memory. From the plurality of straight lines that are candidates for the four edge straight lines detected by using the Hough transform and the plurality of straight lines that are candidates for the upper edge straight lines detected by the upper edge candidate means, the lower side, the left side, It is characterized by detecting a plurality of straight lines that are candidates for edge straight lines on the right side.

  The invention according to claim 17 relates to the image inclination correction method according to any one of claims 11 to 16, wherein the edge selection means determines all combinations of four sides having the same angle between two adjacent sides. Among the combinations of all detected four sides, the combination of the four sides that maximizes the sum of the distance between the upper side and the lower side and the distance between the left side and the right side is regarded as a valid four-side combination as the above-mentioned form. It is characterized by selection.

  The invention according to claim 18 relates to the image inclination correction method according to any one of claims 11 to 17, wherein the corner detecting means is configured to select four sides selected as a combination of four sides that are valid as the form. The feature is that corner coordinates of a form are detected as an intersection of straight lines (including a case where the straight line is an extension) of two adjacent sides in the edge straight line.

  The invention according to claim 19 relates to the image tilt correction method according to any one of claims 11 to 18, wherein the parameter generation means moves the input image coordinates to a specific position and enlarges the image at a specific magnification. Alternatively, after reducing and rotating by a specific angle, the parameter is generated by an inverse matrix of a transformation matrix for obtaining an output image by performing an operation for adjusting the rotation center to the origin of the coordinates of the output image.

  The invention according to claim 20 relates to the image tilt correction method according to any one of claims 11 to 19, wherein the image correction means corrects the tilt of the input form image using the parameter, The correction is performed to remove the outer portion of the form image.

  In the image tilt correction apparatus and the image tilt correction method of the present invention, a scanner device having a paper feed mechanism does not require a search in the sub-scanning direction, and performs processing simultaneously with the input of a raster image using the main scanning direction as a processing unit. Therefore, high-speed processing is possible, and even when the inclination of the form image is small, the corner coordinates of the form image can be obtained with high accuracy and the inclination of the form image can be corrected. In addition, even if the form corner is bent or cut diagonally or rounded, or even if there is a solid area with a small difference in brightness between the outside of the form image and part of the edge of the form, the form image is correct. It is possible to detect the outer edge.

  Using the image read from the line buffer that temporarily stores the input form image for each scan line, the position of the edge points on the top and bottom sides of the form, and the position of the edge points on the left and right sides of the form Using the edge position generation unit for generating the edge position, the edge position storage memory for sequentially storing the generated edge position information, and the edge position information stored in the edge position storage memory, the edges of the four sides of the form An edge candidate detection unit that detects a plurality of straight lines as a straight line candidate, and an edge selection unit that selects a combination of four sides that are valid as a form from a combination of four sides generated from the detected plurality of straight lines. The corner detection unit that detects the corner coordinates of the form from the combination of the four sides, and the parameters required when correcting the form image are generated using the detected corner coordinates. An image storage unit that sequentially stores line buffer information and stores the image as an image, and an image correction unit that corrects the image stored in the image storage memory using the generated parameter. An inclination correction apparatus is configured.

  FIG. 1 is a diagram for conceptually explaining the correction of a form image by the image tilt correction apparatus of the present invention, FIG. 2 is a block diagram showing the configuration of the tilt correction apparatus according to the first embodiment of the present invention, and FIG. FIG. 4 is a flowchart showing the processing procedure of the upper side and lower side edge position generation unit, and FIG. 5 is a processing method for generating the edge positions of the upper side and the lower side. 6 is a diagram illustrating the format of the edge positions of the four sides, FIG. 7 is a diagram for explaining the detection of the edge positions of the upper side and the lower side, and FIG. 8 is the left side and the right side FIG. 9 is a flowchart showing the processing procedure of the edge candidate detection unit, FIG. 10 is a diagram for explaining a method for removing an illegal left edge position and right edge position, and FIG. 11 is FIG. 12 is a flowchart for explaining the selection method of the edge selector, FIG. 13 is a diagram for explaining the corner coordinates detected by the corner detector, and FIG. 14 is an output from the input image. It is a figure explaining the method to calculate an image.

FIG. 1 conceptually illustrates the correction of a form image realized by the image tilt correction apparatus of the present invention.
When a form is read by a scanner having a paper feed mechanism, for example, when the form is set inclined with respect to the normal input position of the scanner, as shown by form PQRS in FIG. The form image is inclined with respect to the outer edge of the input image. Or, even if the form is set correctly, even if the paper is tilted and sent to the scanner due to a problem with the paper feed mechanism of the scanner, the form image is also similar to the outer edge of the input image. It will be inclined.
In such a case, the image inclination correction apparatus of the present invention corrects the output form image so that there is no inclination with respect to the input image by rotating the form image with respect to the input image, for example, around the point P. In addition, the area outside the output form image can be removed, and an image composed only of the area of width W and height H corresponding to the form part can be cut out as shown in FIG. It is what you want to do.

  Next, the configuration of the tilt correction apparatus according to the present embodiment will be described with reference to FIG. The tilt correction apparatus of this example includes a form image input unit 201 such as a scanner having a paper feed mechanism, a line buffer 202 that temporarily stores a raster image input for each scanning line from the form image input unit 201, and An edge position generation unit 203 that detects information on edge points of the upper side, the lower side, the left side, and the right side from the raster image temporarily stored in the line buffer 202, and the raster image temporarily stored in the line buffer 202 are sequentially An image storage memory 204 to store, an edge position storage memory 210 to store edge point information (edge position) generated by the edge position generation unit 203, and each edge from the edge positions stored in the edge position storage memory 210. An edge candidate detection unit 205 that detects a plurality of edge straight line candidates as straight lines that may be configured, and an edge candidate detection unit 20 Edge that detects a reasonable combination as the edge of the form (outer edge) using the straight line slope from the edge straight line candidates detected in step 2 and the distance information of the two opposing straight lines (upper and lower sides, left and right sides) The selection unit 206, the corner detection unit 207 for calculating the corner coordinates of the form by calculating the coordinates of the intersection of two adjacent sides in the straight line of the four sides detected by the edge selection unit 206, and detected by the corner detection unit 207 A parameter generation unit 208 that calculates parameters necessary for image correction from the corner coordinates, and a parameter calculated by the parameter generation unit 208 to correct the image stored in the image storage memory 204 to generate a form And an image correction unit 209 that performs processing for removing the outer portion of the form.

Hereinafter, the function of each component in the tilt correction apparatus of this example shown in FIG. 2 will be described in detail.
A form image to be processed is sequentially input to the form image input unit 201 as a raster image for each scanning line from a scanner having a paper feed mechanism. In this case, it is assumed that the luminance of the outer portion of the form is acquired lower than the luminance of the background portion where no characters or images exist in the form surface.
The line buffer 202 temporarily stores the image data input by the form image input unit 201 for each scanning line.

The edge position generation unit 203 generates the edge positions of the upper side and the lower side from the image data for each line stored in the line buffer 202 by the upper side / lower side edge position generation unit 203a, and the left side / right side edge position generation unit 203b. Thus, the edge positions of the left side and the right side are generated and stored in the edge position storage memory 210. At this time, the edge position generation unit 203 generates an edge position in the format of FIG. Specific operations of the upper side / lower side edge position generation unit 203a and the left side / right side edge position generation unit 203b will be described later.
The image storage memory 204 stores image data output from the line buffer 202 after the edge position generation unit 203 generates the edge position of each side based on the image data temporarily stored in the line buffer 202. Store sequentially.

The edge candidate detection unit 205 detects edge straight line candidates of each side by known Hough transformation from the information on the edge positions of the upper side, lower side, left side, and right side read from the edge position storage memory 210. The edge straight lines of each side are not uniquely determined at this time, and when there are a plurality of edge straight lines with recognized linearity, all of them are generated as candidates. The Hough conversion method will be described later. The detected plurality of edge straight lines are output to the edge selection unit 206.
The edge selection unit 206 selects a plurality of combinations of straight lines having the same (substantially right angle) between two adjacent sides from the edge straight lines detected by the edge candidate detection unit 205 as straight lines constituting four sides, and Among them, a combination that maximizes the sum of the distance between the upper side and the lower side and the distance between the left side and the right side is selected as an appropriate combination as an edge (outer edge) of the form, and is output to the corner detection unit 207.

The corner detection unit 207 detects the corner coordinates of the form by using the edge straight line of each side output from the edge selection unit 206 to obtain the intersection of each straight line. The detected corner coordinates are output to the parameter generation unit 208.
The parameter generation unit 208 generates parameters necessary for image correction based on the output of the corner coordinates from the corner detection unit 207, and outputs the parameters to the image correction unit 209.
The image correction unit 209 corrects the image data stored in the image storage memory 204 using the parameters generated by the parameter generation unit 208, and outputs a corrected form image.

Next, the operation of the tilt correction apparatus of this example will be sequentially described with reference to FIGS.
First, as shown in the flowchart of FIG. 3, image data for the k-th one line is input from the image input unit 201 and temporarily stored in the line buffer 202 (step 301).
Next, the edge position generation unit 203 generates the edge positions of the upper side and the lower side from the image data for one line stored in the line buffer 202 (step 302), and generates the edge positions of the left side and the right side (step 302). 303).

Next, the operation of the upper side / lower side edge position generation unit 203a shown in FIG. 2 will be described with reference to the flowchart of FIG. 4 and FIGS.
The input image data of the kth line is divided into M ranges at s [pixel] intervals in the main scanning direction, and processing is performed using each range as a subsequent processing unit (step 401).
Next, as shown in FIG. 5, the number of white pixels Ai having a luminance T or more is detected from the range i (0 ≦ i ≦ M) (step 402).

Next, if the number of white pixels Ai detected in the range i is s / 2 or more, the process proceeds to step 404, and if it is less than s / 2, the process returns to step 402 (step 403), thereby sequentially The number of white pixels in the range is detected.
Next, in step 404, if the range i first satisfies the condition of step 403 (Ai ≥ s / 2) in the range i, the current edge of the table 601 in FIG. If the number (K) of the image data is set, otherwise, the number (K) of the current image data is set at the location indicating the lower edge position and the range i in the table 601 of FIG. 6 (step 405). ). Table 601 indicates that K is set as the Y coordinate of the edge positions of the upper side and the lower side at the X point of the range i.

Thereafter, as shown in step 407, when the processing of steps 404, 405, and 406 is not completed for the entire range, the process returns to step 402, and when completed for the entire range, the processing is terminated.
By performing such processing for all lines, the edge positions of the upper side and the lower side of the form are generated.

  FIG. 7 illustrates the detection of the upper edge position and the lower edge position. First, the form image is temporarily stored in the line buffer sequentially in units of one scanning line, and then temporarily stored in the line buffer. Is divided into a plurality of ranges at equal intervals, such as range 1, range 2,..., Range M for each predetermined threshold in the main scanning direction, for example, s [pixel], and within each range. At the first line position where the number of existing white pixels is greater than or equal to a predetermined threshold, for example, s / 2 [pixel], information on the edge position of the upper side of the form (●) is generated at the center position of the range. At the last line position where the number of white pixels existing in the range is equal to or greater than the predetermined threshold s / 2 [pixel], information (▼) on the edge position of the lower side of the form is generated at the center position of the range. It is shown.

In the technique disclosed in Patent Document 1, for each line, a change point that changes from a black pixel to a white pixel and a change point that changes from a white pixel to a black pixel are obtained. Then, the number of detected edge points decreases, and in an extreme case, when there is no inclination, the number of detected edge points is “0”.
On the other hand, in the case of the present invention, the form is divided into a plurality of ranges in the main scanning direction, white pixels existing in each range are detected, and the number of detected white pixels exceeds a predetermined threshold value. Since the midpoint within the range is detected as an edge point, the position of the detected edge point and the number of edge points are the same regardless of the magnitude of the inclination. Therefore, in the case of the present invention, even when the inclination of the form image is small, it is possible to accurately detect the upper and lower edge points of the form image.

Next, the operation of the left / right edge position generation unit 203b will be described with reference to the flowchart of FIG. 8 and FIG.
First, the input image data of the k-th line is searched from the left, and the first white pixel position L in which white pixels continue for w [pixel] or more is detected first (step 801). Here, w [pixel] is the number of pixels existing in the main scanning direction within the form width.
If L exists, the process proceeds to step 803, and if L does not exist, the process ends (step 802).

In step 803, the input image data of the Kth line is searched from the left, and finally the last white pixel position R in which white pixels continue for w [pixel] or more is detected.
If R exists, the process proceeds to step 805, and if R does not exist, the process ends (step 804).
In step 805, the position L and the number of the image data at that time are stored as the left edge position in the position corresponding to the current image data number in the table 602 of FIG. 6, and the process proceeds to step 806.

In step 806, the position R and the number of the image data at that time are stored as the right edge position in a location corresponding to the current image data number in the table 602 of FIG.
By performing such processing for all lines of the image data, the edge positions of the left side and the right side are generated.

Next, the edge candidate detection unit 205 detects edge straight line candidates for each side from the edge positions stored in the edge position storage memory 210 according to the flowchart shown in FIG. 9 (step 305). The edge straight lines of each side are not uniquely determined at this time, and when there are a large number of edge straight lines with recognized linearity, all of them are generated as edge straight line candidates. A Hough transform (see Non-Patent Document 1), which is a known technique, is used to detect a straight line that should be an edge straight line candidate.
Hereinafter, the straight line detection operation will be described with reference to the flowchart of FIG. 9 and FIG.

First, information on the edge position of the upper side is voted on the Hough space (step 901). The Hough space used here is a straight line expression (1) shown below.
ρ = X cos θ + Y sin θ (1)
(Ρ, θ) plane transformed from. Here, ρ is the distance from the origin of the coordinates (X = 0, Y = 0) to the straight line, and θ is the angle formed between the perpendicular drawn from the origin to the straight line and the X axis.

Next, all positions in the Hough space that are equal to or greater than the number of votes P are generated as edge straight line candidates on the upper side (step 902).
Further, the information on the edge position of the lower side is voted on the Hough space (step 903), and all positions in the Hough space that are equal to or larger than the number of votes P are generated as the edge straight line candidates on the lower side (step 904).

Next, correct left side edge position information is detected from the information on the edge position on the upper side and the information on the edge position on the lower side (step 905).
This is because, as indicated by 1001 in FIG. 10, the edge position information on the left side includes incorrect left side edge position information that should belong to the lower side.
In order to exclude this, as indicated by 1002 in FIG. 10, an edge point having the maximum Y coordinate among the edge positions of the upper side and an edge point having the minimum Y coordinate among the edge positions of the lower side are represented. Only the edge positions within the range of the edge positions 1003 and 1004 with the margin added are detected as the correct left side edge positions as the reference points.
Then, the correct left side edge position is voted on the Hough space (step 906), and all positions where the number of votes in the Hough space is equal to or greater than P are obtained, and these positions are generated as edge straight line candidates on the left side ( Step 907).

Next, as in the case of the left side, the right side edge position information is detected from the information of the upper side edge position and the lower side edge position (step 908).
Then, the correct right-side edge position is voted on the Hough space (step 909), and all positions where the number of votes in the Hough space is equal to or greater than P are obtained, and these positions are generated as edge straight line candidates on the right side ( Step 910).
In this way, edge straight line candidates for each side are generated.

Next, in the edge selection unit 206, the edge of the form is detected by the four sides having the same angle (substantially a right angle) at which the two adjacent sides meet from the result of the edge straight line candidate detection (step 305) in the edge candidate detection unit 205. A combination of straight lines is selected (step 306).
Then, in the combination of edge straight lines of all the selected forms, a combination of four sides having a validity is detected by the combination of the four sides that maximizes the distance between the upper side and the lower side and the distance between the left side and the right side ( Step 307) When there is a valid combination, the process proceeds to detection of a form corner (Step 308). When there is no valid combination, error processing is performed (Step 311), and the process is performed. finish.

Hereinafter, a method for selecting a combination of edge straight lines in the edge selection unit 206 will be described with reference to the flowchart shown in FIG. 11 and FIG.
First, a variable Max is provided and initialized with 0 (step 1101). Next, all combinations of four sides with the same angle (error ± 0.5 °) are detected (step 1102). However, since the left side and the right side are perpendicular to the upper side and the lower side, adjustment is performed by subtracting the inclination angle from 90 °.

Next, it is checked whether or not a combination of four sides exists (step 1103). If there is no combination, error processing is performed (step 1109), and the process is terminated.
If there is a combination of four sides, as shown at 1201 in FIG. 12, the distance D1 between the upper side and the lower side is calculated (step 1104), and similarly, the distance D2 between the left side and the right side is calculated (step 1105).
Here, the distances D1 and D2 are calculated as shown by 1601 in FIG. 12, when there is a solid area, the value including the solid area is D1, and the value not including the solid area is D2. And

Next, the variable Max is compared with the value of (D1 + D2) (step 1106). If the value of Max is small, information on the combination of the four sides at this time is held, and the variable Max is set to (D1 + D2). A value is substituted (step 1107), and if the value of the variable Max is not small, the process proceeds to step 1108.
In step 1108, it is determined whether or not the processing for all combinations has been completed. If not completed, the process returns to step 1104, and if all the combinations have been completed, the process is terminated.
As a result, as shown at 1202 in FIG. 12, an edge straight line on each side is selected as the outer edge of the form including the solid area.

Next, the corner detection unit 207 uses the edge straight lines of the respective sides selected by the edge selection unit 206 to obtain the intersections of the respective straight lines, and as shown in FIG. The corner coordinates of the four points on the lower left and lower right are detected (step 308).
At this time, since the inclination correction device of the present invention detects the corner coordinates as the intersection of the edge straight lines of each side, the corner coordinates are obtained as the intersection of the extension of the edge straight lines even when a part of the edge straight line is missing. Can do. Therefore, the outer edge of the form image can be correctly detected even when the form corner is bent or cut obliquely or rounded.

Next, the parameter generation unit 208 performs a predetermined calculation based on the four corner coordinates detected by the corner detection unit 207 to generate parameters used by the image correction unit 209 (step 309).
At this time, the conversion formula used in the image correction unit 209 is expressed by the following formula (2), where X is the input image coordinate, U is the output image coordinate, and A is the conversion matrix.
X = A ⁻¹ U (2)
The input image coordinate X, the output image coordinate U, and the transformation matrix A in equation (2) are as follows.

mx : x方向の移動量, sx : x方向の倍率, cx : 回転中心の x座標
my : y方向の移動量, sy : y方向の倍率, cx : 回転中心の y座標
θ : 回転角度

mx: amount of movement in x direction, sx: magnification in x direction, cx: x coordinate of rotation center
my: displacement in y direction, sy: magnification in y direction, cx: y coordinate of rotation center θ: rotation angle

In the above equations, x represents the coordinate in the main scanning direction, and y represents the coordinate in the sub scanning direction. In the conversion matrix A, the first term indicates the amount of movement of the form image to a specific position. The second term indicates the magnification of the form image. The third term indicates the rotation angle of the form image. The fourth term indicates that the center of rotation is aligned with the origin (0, 0) of the x and y coordinates.
The enlargement / reduction of the form image is to compensate for the expansion / contraction of the form image in the x direction and the y direction due to the characteristics of the scanner. When the expansion / contraction amount is sufficiently small, the magnification (sx, sy) = ( 1,1).
In this case, the inverse matrix A ⁻¹ of the transformation matrix A is expressed by the following equation (3).

Therefore, in the parameter generation unit 208, the six parameters a ₁₁ , a ₁₂ , a ₁₃ , a ₂₁ , a from the mx, my, sx, sy, cx, cy, θ in the expressions (2) and (3). _22, thereby obtaining the _{a 23.}
As an example, when the inclination of the form is corrected as shown in FIG. 1 and unnecessary portions outside the form are removed, the rotation angle θ can be obtained from the inclinations of the four-side straight lines detected by the edge selection unit 206. If the center of rotation (cx, cy) is the upper left corner coordinate position, the movement amount (mx, my) = (0, 0), and the magnification (sx, sy) = (1, 1) and the parameter a _{11, a 12, a 13,} a 21, a 22, a 23 is calculated.

Finally, the image correction unit 209 corrects the inclination of the form by using the parameters generated by the parameter generation unit 208, and performs correction to remove unnecessary portions outside the form (step 310).
When correction is performed as shown in FIG. 1, an output image is generated for a range indicated by the following expression (4) of the output image coordinates U using Expression (2).
At this time, the value of each pixel of the output image can be smoothly corrected by performing conversion by applying a known interpolation technique such as a linear interpolation method.
0 ≦ u ≦ W, 0 ≦ v ≦ H (4)
Here, the output image U = (u, v), u is the output coordinate in the X direction, and v is the output coordinate in the Y direction. In Equation (4), W is the width of the form, and H is the height of the form.
Thereby, an unnecessary area outside the form image is removed. Note that the process of correcting the inclination of the form image and the process of removing an unnecessary area outside the form image are performed in parallel in the image correcting unit 209.

Hereinafter, a method for calculating an output image from an input image will be described in more detail with reference to FIG.
Now, on the input image in the coordinate system x, y shown in FIG. 14 (a), when the form exists in the area indicated by P, Q, R, S, in the coordinate system u, v shown in FIG. 14 (b). Assume that the output image indicating the form occupies the areas indicated by A, B, C, and D.
The above formula (4) represents an area surrounded by A, B, C, and D, and this area is a form part excluding the outside of the form. That is, the output image is generated as having a width of W and a height (height) of H. The horizontal width W and the vertical width H can be calculated from four form corners.

More specifically, the coordinates of the input image corresponding to the coordinates A (0,0) of the origin of the output image are calculated by substituting the coordinates (0,0) of A into the equation (2). The coordinate value at that time is point P in FIG. Similarly, for B, C, and D, the coordinates in the corresponding input image can be calculated from Equation (2), and the points are Q, R, and S in FIG.
If such an operation is performed on all coordinates (W × H) included in the region surrounded by A, B, C, and D of the output image, as shown in FIG. An output image (corrected image) in which the inclination is corrected and the outside of the form is removed is generated.

At this time, the coordinates in the input image calculated by Expression (2) are coordinates including a decimal point. On the other hand, since the input image is a digital image and is discretized, the coordinates of the input image are integers. Therefore, an error occurs even if the coordinates of the output image and the coordinates of the input image are to be associated with each other.
The simplest method for avoiding such an error is to round the numbers after the decimal point to make the luminance of the coordinates on the input image side converted to an integer the luminance on the output image side. There is a tendency that the vicinity of the edges becomes jagged. On the other hand, smooth interpolation can be performed by performing interpolation using an interpolation technique such as a known linear interpolation method.

  As described above, according to the tilt correction apparatus of this example, the scanner apparatus having the paper feed mechanism does not require a search in the sub-scanning direction, and at the same time as the input of the raster image with the main scanning direction as a processing unit. Since processing can be performed, high-speed processing is possible, and even when the inclination of the form image is small, the corner coordinates of the form image can be obtained with high accuracy and the inclination of the form image can be corrected. Furthermore, there is no need to detect a stable area, and there is little difference in brightness between the outside of the form image when the form corner is bent or slanted or rounded, or part of the form edge. Even when there is a solid area, the outer edge of the form image can be detected correctly.

FIG. 15 is a block diagram showing a configuration of a tilt correction apparatus according to the second embodiment of the present invention, and FIG. 16 is a flowchart showing a processing procedure in the tilt correction apparatus of the present embodiment.
The configuration of the inclination correction apparatus of this example is substantially the same as that of the inclination correction apparatus of the first embodiment shown in FIG. 2, but instead of the edge candidate detection unit 205, a lower side / left side / right side edge candidate detection unit 212 is provided. In addition, a difference is that an upper edge candidate detection unit 211 is additionally provided between the edge position storage memory 210 and the lower side / left side / right side edge candidate detection unit 212.
The processing procedure of the inclination correction apparatus of this example is substantially the same as that of the inclination correction apparatus of the first embodiment shown in FIG. 3, but has step 312 between step 303 and step 304. A difference is that step 314 is provided instead of step 305 between step 304 and step 306, and step 313 is provided between step 312 and step 314.

Hereinafter, in the configuration of the inclination correction apparatus of this example, a part different from the case of the first embodiment will be described in detail.
The upper edge candidate detection unit 211 detects the information on the edge straight line of the upper side generated by the edge position generation unit 203 via the edge position storage memory 210 and transmits the information to the lower side / left side / right side edge candidate detection unit 212. .
The lower side / left side / right side edge candidate detection unit 212 detects edge straight line candidates for the lower side, left side, and right side using the information on the edge positions of the lower side, left side, and right side stored in the edge position storage memory 210, and The information is transmitted to the edge selection unit 206 together with information on the edge straight line of the upper side detected by the upper edge candidate detection unit 211.

Next, the operation of the tilt correction apparatus of this example will be described. The operations of steps 301 to 310 in the tilt correction apparatus of this example are the same as those in the first embodiment. In the following, portions different from the case of the first embodiment will be described in detail.
In step 312, when the number G of the input image data (indicated by the number of input lines) becomes G by predetermining the number G of the image data that the upper side of the form is necessarily acquired, the edge straight line on the upper side. Detection is triggered. Note that the value of G does not need to be a very large value, and may be a number of image data within a range where the inclination is about 10 °, for example, according to the inclination that may actually occur.
In step 313, the upper edge straight line candidate is detected in advance using the upper edge position generated at the present time.
In step 314, after the edge straight line candidate on the upper side in step 313 is detected, the edge straight line candidates on the lower side, the left side, and the right side are detected.

  In the tilt correction apparatus of the first embodiment described above, edge straight line candidates are detected after generating edge positions for all lines of image data. In the tilt correction apparatus of this example, generation of the edge positions other than the upper side is not completed until all the image data lines are input, but the upper edge position is completed when a certain amount of image data lines are input. Since the detection of the upper edge straight line candidate is performed in advance at the time when the generation of the upper edge position is completed, the entire process is compared with the case of the first embodiment. Time can be shortened.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and the present invention can be changed even if there is a design change or the like without departing from the gist of the present invention. include. For example, the size of the form to be tilt-corrected, the vertical / horizontal shape ratio, and the like are arbitrary.

  The present invention may be a scanner device having a paper feed mechanism in which a form is moved with respect to an optical sensor as in a facsimile machine, or a form in which a form is fixed and the optical sensor is moved as in a copying machine. Can also be used.

It is a figure which illustrates notionally correction | amendment of the form image by the image inclination correction apparatus of this invention. It is a block diagram which shows the structure of the inclination correction apparatus which is 1st Example of this invention. It is a flowchart which shows the process sequence in the inclination correction apparatus of the Example. It is a flowchart which shows the process sequence of an upper side and a lower side edge position production | generation part. It is a figure explaining the processing method at the time of producing | generating the edge position of an upper side and a lower side. It is a figure which illustrates the format of the edge position of 4 sides. It is a figure for demonstrating the detection of the edge position of an upper side, and the edge position of a lower side. It is a flowchart which shows the process sequence of the left side and right side edge position generation part. It is a flowchart which shows the process sequence of an edge candidate detection part. It is a figure explaining the method of removing an illegal left side edge position and a right side edge position. It is a flowchart which shows the process sequence of an edge selection part. It is a figure which illustrates the selection method of an edge selection part. It is a figure which illustrates the corner coordinate detected by a corner detection part. It is a figure explaining the method of calculating an output image from an input image. It is a block diagram which shows the structure of the inclination correction apparatus which is 2nd Example of this invention. It is a flowchart which shows the process sequence of the inclination correction apparatus of the Example. It is a figure which illustrates solid coating of the form edge vicinity.

Explanation of symbols

201 Form image input unit (form image input means)
202 Line buffer 203 Edge position generator (edge position generator)
203a Upper / lower edge position generation unit 203b Left / right edge position generation unit 204 Image storage memory 205 Edge candidate detection unit (edge candidate detection means)
206 Edge selection unit (edge selection means)
207 Corner detection unit (corner detection means)
208 Parameter generator (parameter generator)
209 Image correction unit (image correction means)
210 Edge position storage memory 211 Upper edge candidate detection unit (upper edge candidate detection means)
212 Lower / Left / Right Edge Candidate Detection Unit (Lower / Left / Right Edge Candidate Detection Unit)

Claims (20)

  Using the image read from the line buffer that temporarily stores the input form image for each scan line, the position of the edge points on the top and bottom sides of the form, and the position of the edge points on the left and right sides of the form Using the edge position generation unit for generating the edge position, the edge position storage memory for sequentially storing the generated edge position information, and the edge position information stored in the edge position storage memory. An edge candidate detection unit that detects a plurality of straight lines as edge straight line candidates, and an edge selection unit that selects a combination of four sides that are valid as a form from a combination of four sides generated from the detected plurality of straight lines And a corner detection unit that detects corner coordinates of the form from the combination of the selected four sides, and is necessary when correcting the form image using the detected corner coordinates. A parameter generation unit that generates parameters to be used, an image storage memory that sequentially stores the information of the line buffer and stores it as an image, and corrects the image stored in the image storage memory using the generated parameter An image tilt correction apparatus comprising: an image correction unit.   Using the image read from the line buffer that temporarily stores the input form image for each scan line, the position of the edge points on the top and bottom sides of the form, and the position of the edge points on the left and right sides of the form An edge position generation unit that generates the edge position storage memory that sequentially stores the generated edge position information, and the edge position information stored in the edge position storage memory. An upper edge candidate detection unit that detects a plurality of straight lines as edge straight line candidates, edge position information stored in the edge position storage memory, and an upper edge straight line detected by the upper edge candidate detection unit. From the information, a lower side, left side, right side edge candidate detecting unit for detecting a plurality of straight lines as edge straight line candidates forming the lower side, left side, and right side of the form, and the detected complex An edge selection unit that selects a combination of four sides that is appropriate as a form from a combination of four sides generated from the straight line, a corner detection unit that detects corner coordinates of the form from the selected combination of four sides, A parameter generation unit that generates parameters necessary for correcting the form image using the detected corner coordinates, an image storage memory that sequentially stores the information of the line buffer and saves it as an image, and the generated parameters An image tilt correction apparatus comprising: an image correction unit that corrects an image stored in the image storage memory.   The edge position generation unit divides the image data stored in the line buffer into a plurality of ranges at equal intervals, and the number of pixels in the form portion existing in each range is ½ of the number of pixels in the range. When it is detected for the first time, the process of setting the image data number as the position of the upper edge point at the midpoint of the range is performed for each line, and the number of pixels of the form part existing in each range is determined. Upper side and lower side for performing processing for setting the image data number for each line when the middle point of the range is detected as the position of the lower side edge point when it is finally detected that the number of pixels in the range is ½ or more The edge position generator and the image data are searched in the line direction, the position of the left edge point is detected by the first pixel position where the pixels of the form part continue beyond the form width, and the position of the right edge point by the last pixel position Inspect A left side / right side edge position generation unit that performs processing for each line, and configured to store information on the generated edge positions of the four sides in the edge position storage memory. The image tilt correction apparatus according to 1 or 2.   The edge candidate detection unit is configured to detect a plurality of straight lines that are candidates for four-sided edge straight lines using Hough transform from the positions of the four-sided edge positions stored in the edge position storage memory. The image tilt correction apparatus according to claim 1 or 3, wherein   The upper edge candidate detection unit is configured to detect a plurality of straight lines that are candidates for the upper edge straight line using Hough transform from the information on the edge position of the upper edge stored in the edge position storage memory. The image inclination correction apparatus according to claim 2 or 3, characterized by the above-mentioned.   The lower side, left side, and right side edge candidate detection unit includes a plurality of straight lines that are candidates for four-sided edge straight lines detected by using Hough transform from the information on the four-sided edge positions stored in the edge position storage memory, and And a plurality of straight lines that are candidates for edge straight lines on the lower side, the left side, and the right side are detected from a plurality of straight lines that are candidates for the upper side edge straight line detected by the upper side edge candidate detection unit. The image inclination correction apparatus according to claim 5.   The edge selection unit detects all combinations of four sides having the same angle between two adjacent sides, and among all the detected combinations of four sides, the distance between the upper side and the lower side, the distance between the left side and the right side, 7. The image inclination correction device according to claim 1, wherein a combination of four sides that maximizes the sum of the four sides is selected as a combination of four sides that is appropriate as the form. .   The corner detection unit is an intersection of straight lines (including the case of an extension of a straight line) that forms two adjacent sides in a four-sided edge straight line selected as a combination of four sides that are valid as the form. 8. The image tilt correction apparatus according to claim 1, wherein the image tilt correction apparatus is configured to detect corner coordinates.   The parameter generation unit moves the coordinates of the input image to a specific position, enlarges or reduces the image by a specific magnification, rotates the image by a specific angle, and then performs an operation for adjusting the rotation center to the origin of the coordinates of the output image. 9. The image tilt correction apparatus according to claim 1, wherein the parameter is generated by an inverse matrix of a transformation matrix for obtaining an output image coordinate.   10. The image correcting unit is configured to correct an inclination of an input form image and to perform correction for removing an outer portion of the form image using the parameter. The image inclination correction apparatus according to any one of the above.   Using the image read from the line buffer that temporarily stores the input form image for each scanning line, the edge position generation means uses the positions of the edge points of the upper and lower sides of the form, the left side of the form, and The edge position information on the right side is generated, the generated edge position information is sequentially stored in the edge position storage memory, and the edge candidate detection means uses the edge position information stored in the edge position storage memory. Then, a plurality of straight lines are detected as candidates for the four edge straight lines of the form, and the edge selection means detects the four valid sides as a form from the combination of the four sides generated from the detected plurality of straight lines. The combination is selected, the corner detection means detects the corner coordinates of the form from the selected combination of the four sides, and the parameter generation means detects the detected corner. An image in an image storage memory in which parameters necessary for correcting a form image are generated using coordinates, and image correction means sequentially stores the information of the line buffer using the generated parameters and stores the information as an image. An image inclination correction method characterized by correcting the image.   Using the image read from the line buffer that temporarily stores the input form image for each scanning line, the edge position generation means uses the positions of the edge points of the upper and lower sides of the form, the left side of the form, and And the edge position information on the right side is generated, the generated edge position information is sequentially stored in the edge position storage memory, and the upper edge candidate detection means stores the edge position information stored in the edge position storage memory. A plurality of straight lines are detected as edge straight line candidates forming the upper side of the form, and the lower side, left side, and right side edge candidate detecting means includes edge position information stored in the edge position storage memory, and the upper side From the edge straight line information detected by the edge candidate detection means, a plurality of straight lines are detected as edge straight line candidates forming the lower, left and right sides of the form, and the edges are detected. A selection unit selects a combination of four sides that is appropriate as a form from a combination of four sides generated from the detected plurality of straight lines, and a corner detection unit selects a corner of the form from the combination of the selected four sides. The coordinates are detected, the parameter generation means generates parameters necessary for correcting the form image using the detected corner coordinates, and the image correction means uses the generated parameters to generate the parameters of the line buffer. An image tilt correction method comprising correcting an image stored in an image storage memory in which information is sequentially stored and stored as an image.   In the edge position generation means, the upper and lower edge position generation means divides the image data stored in the line buffer into a plurality of ranges at equal intervals, and the number of pixels of the form part existing in each range is When it is first detected that the number of pixels in the range is ½ or more, the process of setting the number of the image data is performed for each line with the midpoint of the range as the position of the upper edge point. When the number of pixels of the form part existing in the last is detected to be 1/2 or more of the number of pixels in the range, the image data number is set with the middle point of the range as the position of the lower edge point. The processing is performed for each line, and the left side / right side edge position generation means searches the image data in the line direction, detects the position of the left side edge point from the first pixel position where the pixels of the form portion continue for the form width or more, 13. The processing for detecting the position of the right edge point based on the pixel position is performed for each line, and the generated information on the edge positions of the four sides is stored in the edge position storage memory. Image tilt correction method.   The edge candidate detection means detects a plurality of straight lines that are candidates for four-sided edge straight lines using Hough transform from information on the four-sided edge positions stored in the edge position storage memory. Item 14. The image tilt correction method according to Item 11 or 13.   The upper edge candidate detection means detects a plurality of straight lines that are candidates for an upper edge straight line by using Hough transform from information on the edge position of the upper edge stored in the edge position storage memory. 14. An image tilt correction method according to 12 or 13.   The lower side, left side, and right side edge candidate detecting means includes a plurality of straight lines that are candidates for edge straight lines of four sides detected by using Hough transform from the information on the edge positions of the four sides stored in the edge position storage memory; 16. A plurality of straight lines that are candidates for edge straight lines on the lower side, the left side, and the right side are detected from a plurality of straight lines that are candidates for the upper side edge straight lines detected by the upper side edge candidate means. The image inclination correction method described.   The edge selection means detects all combinations of four sides having the same angle formed by two adjacent sides, and among all the detected combinations of four sides, the distance between the upper side and the lower side, the distance between the left side and the right side, 17. The image inclination correction method according to claim 11, wherein a combination of four sides that maximizes the sum of the values is selected as a combination of four sides that is appropriate as the form.   The corner detection means uses the form of the form as an intersection of straight lines (including the case of an extension of a straight line) that constitutes two adjacent sides in the four-sided edge straight line selected as a combination of four sides that are valid as the form. 18. The image inclination correction method according to claim 11, wherein corner coordinates are detected.   The parameter generation means moves the input image coordinates to a specific position, enlarges or reduces the image by a specific magnification, rotates the image by a specific angle, and then performs an operation for adjusting the rotation center to the origin of the coordinates of the output image. 19. The image inclination correction method according to claim 11, wherein the parameter is generated by an inverse matrix of a transformation matrix for obtaining an output image.   The said image correction means correct | amends the inclination of an input form image, and performs the correction | amendment which removes the outer part of a form image using the said parameter. Image tilt correction method.

Images