JP2011039896A - Image processing apparatus and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To derive a feature quantity characterizing an image having a frame from a correctly determined layout of segments constituting the frame even when the image is rotated, translated, scaled or otherwise distorted. <P>SOLUTION: An image processing apparatus 1 includes a segment extraction means for extracting segments included in an image, a segment selection means for selecting segments out of the extracted segments according to a predetermined standard, a feature quantity derivation means for selecting segments substantially parallel to the respective edges of the image out of the selected segments according to the distance from each edge and deriving a feature quantity characterizing the image, a linear map derivation means for deriving a linear map that matches the feature quantity of an image to the feature quantity of another image selected according to the size of an area enclosed by the segments selected by the feature quantity derivation means, and an image comparison means for comparing the latter image and the former image transformed via the linear map derived by the linear map derivation means. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an image processing apparatus and a computer program.

  In Patent Document 1, input image data and an output print image are measured, and an image feature amount based on one of an edge, a maximum point, and a minimum point of the image is calculated for the measured output print image data. An image distortion correction method is described in which, based on the amount, the positional relationship and displacement amount of the corresponding images of the input image data and the output print image data are obtained, and a distortion correction formula is calculated.

  In Patent Document 2, a horizontal black run histogram and a vertical black run histogram are created for the image A and the image B of a form having ruled lines, and the black run histogram for the image A and the black run histogram for the image B are created. Describes a method for detecting a shift of an image by detecting a shift between the two.

  In Patent Document 3, an x-axis projection histogram and a y-axis projection histogram are generated for a reference image and an input image of a form having ruled lines in consideration of the inclination direction of the upper side of the book, and the detected peak values are used for the existence of ruled lines. Pre-print removal that detects the amount of misalignment of the reference image and the input image by associating the reference image and the input image where the difference between the coordinates where the ruled line exists is smaller than a preset threshold value An apparatus is described. The apparatus performs coordinate conversion based on the detected amount of displacement and collates the images.

JP 2001-186337 A JP-A-11-120288 JP 2002-57873 A

  In the present invention, when a plurality of images that are deformed by rotation, parallel movement, enlargement / reduction, etc. are larger than a certain limit, the deformation is removed and one image and another image are removed. The aim is to solve the problem of being unable to compare.

  In order to solve the above-described problem, an image processing apparatus according to claim 1 of the present invention includes a line segment extracting unit that extracts a line segment included in an image, and a predetermined reference among the extracted line segments. A line segment selecting means for selecting a line segment based on the line segment, a line segment that is substantially parallel to each edge of the image is selected based on the distance from each edge of the selected line segment, and the image A feature amount deriving unit for deriving a feature amount that characterizes the image, and a feature amount of one image selected based on a size of an area surrounded by a line segment selected by the feature amount deriving unit Linear mapping deriving means for deriving a linear mapping that matches the image, and image comparing means for comparing the one image with another image converted by the linear mapping derived by the linear mapping deriving means.

  An image processing apparatus according to claim 2 of the present invention is the image processing apparatus according to claim 1, wherein a part of a graphic included in the image is extracted, and the part is newly extracted as the image. Have means.

  An image processing apparatus according to claim 3 of the present invention is the image processing apparatus according to claim 2, wherein the partial extraction unit performs a filling process starting from at least one point near the outer edge or the outer edge of the image. Contour graphic acquisition means for acquiring a contour graphic, which is a graphic indicating the contour of the obtained graphic.

  An image processing apparatus according to a fourth aspect of the present invention is the image processing apparatus according to the second or third aspect, wherein the partial extraction unit identifies a connected graphic that is a connected graphic included in the image. Connected figure identifying means; and maximum connected figure obtaining means for obtaining a maximum connected figure which is the largest connected figure among the connected figures.

  An image processing apparatus according to a fifth aspect of the present invention is the image processing apparatus according to any one of the first to fourth aspects, wherein the feature quantity storage stores the feature quantity obtained for one image. And a line segment extraction range limiting unit that limits a range in which the line segment extraction unit extracts a line segment for the other image based on the feature amount stored in the feature amount storage unit.

  According to a sixth aspect of the present invention, there is provided a computer program comprising: a line segment extracting means for extracting a line segment included in an image; and a line segment based on a predetermined criterion among the extracted line segments. A line segment selecting means for selecting a line segment, a line segment that is substantially parallel to each edge of the image among the selected line segments is selected based on a distance from each edge, and a feature amount that characterizes the image And a linear mapping that matches the feature quantity of one image with the feature quantity of one image selected based on the size of the region surrounded by the line segment selected by the feature quantity derivation means The linear mapping deriving means for deriving the image, and the image comparing means for comparing the one image with another image converted by the linear mapping derived by the linear mapping deriving means.

  According to the image processing apparatus of the first aspect, even when a plurality of images deformed by rotation, parallel movement, enlargement / reduction, and the like are deformed to each other, the deformation is larger than a certain limit. And one image can be compared with another image.

  Further, according to the image processing device of the second aspect, it is possible to reduce the amount of calculation when extracting line segments, compared to an image processing device that does not have the configuration specified in the present claim.

  Further, according to the image processing apparatus of the third aspect, it is possible to exclude the figure inside the frame from the target for extracting the line segment.

  Further, according to the image processing apparatus of the fourth aspect, it is possible to exclude graphics other than the frame from the target for extracting the line segment.

  Further, according to the image processing apparatus of the fifth aspect, when deriving the feature amount of the image having the same kind of frame as the image from which the feature amount has been derived, the image not having the configuration specified in the present claim Compared to the processing apparatus, the amount of calculation when extracting line segments can be reduced.

  Further, according to the computer program according to the sixth aspect, even when a plurality of images deformed with each other by rotation, parallel movement, enlargement / reduction, etc. by the computer, the degree of the deformation is larger than a certain limit. By removing such deformation, one image can be compared with another image.

1 is a functional block diagram of an image processing apparatus that is an example of an embodiment of the present invention. It is a functional block diagram of an image feature-value deriving unit. It is the 1st image received by the image reception part. It is a figure which shows the 1st image after a filling process. It is a figure which shows the 1st image after the outline acquisition. It is a figure which shows the 1st image in which only the largest connection figure was left. It is a figure which shows the 1st image to which only the largest connection figure acquisition process was performed. 7 is an example of a plot on an r-θ plane obtained by Hough transforming the first image of FIG. 6. It is a figure which shows the position of the intersection which is the feature-value of a 1st image when a line segment selection part employ | adopts the 3rd reference | standard. It is a figure which shows a 2nd image. It is a figure which shows the area | region excluded from line segment extraction object. It is the 2nd image by which inclination, a position, and a size were corrected. It is a figure which shows a difference image.

  FIG. 1 is a functional block diagram of an image processing apparatus 1 that is an example of an embodiment of the present invention.

  The image processing apparatus 1 is realized by using a general computer having a CPU (Central Processing Unit), a memory, an input / output interface, a monitor, an external storage device, and the like that are physically general-purpose information processing apparatuses. And the image processing apparatus 1 is virtually implement | achieved by running the computer program for operating a computer as the image processing apparatus 1 on this computer. A computer program for operating the computer as the image processing apparatus 1 is provided by being recorded on an arbitrary information recording medium such as a DVD-ROM (DVD-Read Only Memory) or a CD-ROM (Compact Disk Read Only Memory). Alternatively, it may be provided as a computer-readable electric signal via an electric communication line such as a public line typified by the Internet. Of course, it is not always necessary to use a general-purpose information processing apparatus for realizing the image processing apparatus 1, and even if a dedicated apparatus is manufactured, or as a module mounted or added to office equipment such as a copier or a facsimile machine. It may be configured.

  Note that the functional block diagram of the illustrated image processing apparatus 1 shows the image processing apparatus 1 realized by a computer program in terms of its functions for convenience of explanation, and each functional block is not necessarily included. It does not necessarily exist physically.

  The image processing apparatus 1 includes an image receiving unit 2 that receives an image, a first image storage unit 3 that stores a first image, a second image storage unit 4 that stores a second image, and the received image An image feature quantity deriving unit 5 for deriving image feature quantities, a first feature quantity storage unit 6 and a second feature quantity storage unit 7 for storing the obtained feature quantities, a linear mapping deriving unit 8, and an image conversion unit 9 and an image comparison unit 10 and a difference image output unit 11.

  The image referred to here is so-called raster image data such as a bitmap, and preferably has a frame around its periphery. The frame need only be generally rectangular, and need not be strictly rectangular. The content of such an image is not particularly limited, but typical examples of such an image include a drawing image, a form image, and a data sheet such as MSDS (Material Safety Data Sheet). In the following description of the present embodiment, the image of the drawing will be described as an example, but the present invention is not intended to be limited to this.

  The feature amount is an amount that characterizes the image. More specifically, the feature amount is an amount that indicates a position, a size, and a rotation angle that a graphic included in the image occupies in the image. Although what value is used as a feature amount is not particularly limited, as an example, information indicating the position of a line segment constituting a frame included in an image is a feature amount. Such information may be a value that directly indicates the position or angle at which each line segment is arranged in the image, or may be the coordinates of the intersection point of each line segment, that is, the point that corresponds to the corner of the frame.

  The image receiving unit 2 is an interface that receives electronic image data from the outside. The image receiving unit 2 may be anything as long as it can receive an image. Specifically, the image receiving unit 2 is connected to a WAN (Wide Area Network) including a public line such as a LAN (Local Area Network) or the Internet. A connected telecommunication line, an arbitrary information recording medium reading device, a scanner for reading a paper medium, and the like are conceivable.

  When the first image is received by the image receiving unit 2, it is stored in the first image storage unit 3 and transferred to the image feature quantity deriving unit 5. When the second image is received, the second image is stored in the second image storage unit 4 and transferred to the image feature amount deriving unit 5.

  FIG. 2 is a functional block diagram of the image feature quantity deriving unit 5. Hereinafter, the operation of the image feature quantity deriving unit 5 will be described with reference to specific examples.

  FIG. 3 is a first image 100 received by the image receiving unit 2 (see FIG. 1). The first image 100 is obtained by capturing a drawing printed out on paper as electronic data with a scanner. The first image 100 is drawn in a rectangular frame 101, a title column 102 for entering a drawing title, an approval column 103 for entering an approval mark or signature, an outer frame 104 for a parts table, and the frame 101. The figure 105 etc. are included. Of course, this is an example for explanation, and is not intended to limit the first image to such a specific example. The first image 100 is assumed to have a black background and a frame 101 etc. drawn on a white background, but this is also an example, and black and white may be reversed or different tones may be used.

  Returning to FIG. 2, the first image 100 received by the image receiving unit 2 is transferred from the first image storage unit 3 to the image correction unit 12 and is corrected if necessary. The correction performed here is for smoothly proceeding with image processing described later, and various known corrections may be performed as necessary. These corrections include binarization processing that converts a grayscale or full-color image into a black and white binary image, discontinuity compensation that compensates for discontinuity of line segments in the image due to blurring, and noise that eliminates noise in the image. There are various types such as removal, but in this embodiment, binarization processing is performed. In the binarization process, a threshold value of a certain luminance is determined, and black may be set if the luminance of each pixel in the image is lower than the threshold value, and white if the luminance is higher.

  Subsequently, the first image 100 is passed to the line segment extraction range limiting unit 13 that functions as a line segment extraction range limiting unit that limits the range in which the line segment extraction unit 18 extracts line segments. The operation of the line segment extraction range restriction unit 13 will be described in detail later.

  Further, the first image 100 is delivered to a partial extraction unit 14 that functions as a partial extraction unit that extracts a part of a graphic included in the image and newly uses the part as the image. The partial extraction unit 14 includes a contour graphic acquisition unit 15 that functions as a contour graphic acquisition unit that acquires a contour graphic that is a graphic indicating a contour of a graphic obtained by a filling process starting from at least one point in the vicinity of the outer edge or the outer edge of the image. As a connected figure identifying unit 16 that functions as a connected figure identifying means for identifying a connected figure that is a connected figure included in an image, as a maximum connected figure obtaining means for obtaining a maximum connected figure that is the largest connected figure among the connected figures. It has the maximum connected figure acquisition part 17 which functions.

  First, the first image 100 is transferred to the contour graphic acquisition unit 15. The contour graphic acquisition unit 15 first performs a painting process starting from at least one point around the first image 100. This painting process may be performed by a known method. FIG. 4 shows the first image 106 after the filling process, and the shaded portion in the drawing is an area where the filling process has been performed.

  Then, the contour graphic acquisition unit 15 acquires the contour of the region subjected to the fill processing. A well-known method may be used for acquisition of the contour, and an example thereof is application of a differential filter. FIG. 5 is a diagram showing the first image 107 after the contour acquisition. Compared with FIG. 3, the title column 102 and the graphic 105 inside the frame 100 are excluded, and the outer shape and the outer frame 104 of the frame 100 and the approval column 103 remain. The contour graphic acquisition unit 15 delivers the first graphic 107 after the image processing to the connected graphic identification unit 16. Here, the black and white of the image may be inverted before the outline of the region that has been subjected to the filling process is acquired.

  The connected graphic identification unit 16 classifies and identifies the graphic included in the first image 107 for each connected graphic (that is, for each connected graphic). This process may be a process generally known as labeling in image processing. Specifically, by repeating the operation of giving identification information to a black pixel included in the first image 107 and further giving the same identification information to a black pixel adjacent to the pixel, the first image 107 is obtained. Different identification information is assigned to each connected graphic for all black pixels included in the image 107. As a result, as shown in FIG. 5, the identification information “A” is given to the outer shape of the frame 101 and the approval column 103 in the first image 107 as the connected figure A, and the identification information “B” is displayed in the outer frame 104. "Is given and identified as a connected figure B.

  Subsequently, the maximum connected figure obtaining unit 17 obtains the largest connected figure among the connected figures included in the first image 107. Here, the connected figure A, which is a large connected figure, is selected by comparing the sizes of the connected figure A and the connected figure B, in particular, the width and height thereof. Note that there are various methods for comparing the sizes, and there is no particular limitation. As a method for comparing width and height, there are a method using one of them, a method for comparing both of them and giving priority, a method for comparing using the sum of width and height, and the like. Further, the comparison may be performed using the area of the connected figure or the area of the rectangle circumscribing the connected figure. As a result, as shown in FIG. 6, only the connected graphic A, which is the maximum connected graphic, remains in the first image 108. The maximum connected figure acquisition unit 17 delivers the obtained first image 108 to the line segment extraction unit 18.

  Through the above processing, the connected figure A as a part thereof is extracted from the first image 100 (see FIG. 3). In the present embodiment, the partial extraction unit 14 performs both the contour graphic acquisition process by the contour graphic acquisition unit 15 and the maximum connected graphic acquisition process by the connected graphic identification unit 16 and the maximum connected graphic acquisition unit 17 in this order. However, both of these may not necessarily be performed, and either one may be used.

  First, when only the outline graphic acquisition process by the outline graphic acquisition unit 15 is performed, the first image 107 shown in FIG. 5 is transferred to the line segment extraction unit 18. In addition, when only the maximum connected graphic acquisition process by the connected graphic identification unit 16 and the maximum connected graphic acquisition unit 17 is performed, the above-described maximum connected graphic acquisition process is performed on the first image 100 illustrated in FIG. 3. Become. As a result, the first image 109 shown in FIG. 7 is obtained, and only the connected figure C including the frame 101, the title field 102, and the approval field 103 is included in the image as the largest connected figure.

  Subsequently, the first image 108 (or the first image 107 or 109) is transferred to the line segment extraction unit 18 that functions as a line segment extraction unit that extracts a line segment included in the image.

  The line segment extraction unit 18 extracts a line segment included in the transferred first image 108 and specifies information related to the position and length or length. Various known methods can be used to extract this line segment, and typical methods include Hough transform, projection distribution (histogram), tracking, etc. In this embodiment, noise is extracted. Therefore, the Hough transform is employed because it is not greatly affected by the interruption of the line segment due to the above or the inclination of the connected graphic A with respect to the outer shape of the first image 108. Of course, it is arbitrary to employ a method other than the Hough transform.

The first image 108 is converted to r-θ coordinates by the Hough transform, and the line segment included in the first image 108 is represented as a point with a large number of votes on the r-θ plane. The number of votes is an amount related to the length of the line segment. Generally, the greater the number of votes, the longer the length of the line segment. The method of taking r and θ is arbitrary, but here r represents the length of the normal drawn from the lower left point of the first image 108 with respect to the line segment, and θ represents the angle of the normal. And The range of r is 0 ≦ r ≦ √X ² + Y ² where X is the horizontal length of the first image 108 and Y is the vertical length.
It becomes. Further, it is sufficient that the range of θ is a half circumference, that is, a range of 180 °, but here, as an example,
−45 ° ≦ θ <135 °.

  The obtained result is transferred to a line segment selection unit 19 that functions as a line segment selection unit that selects a line segment based on a predetermined criterion among the extracted line segments. FIG. 8 is an example of a plot on the r-θ plane obtained by Hough transforming the first image 108 of FIG. In the figure, the points indicated by black circles are points indicating line segments obtained by the Hough transform, and the size of the black circles reflects the number of votes, that is, the length of the line segments. Further, alphabets a to f written beside the black circles indicate that they correspond to the respective line segments a to f shown in FIG.

  The line segment sorting unit 19 sorts line segments based on a predetermined criterion among these obtained line segments. There are various ways to determine this standard.

The first standard is the simplest one, and is a standard for selecting the vertical and horizontal line segments having the lengths up to the top nth. On the r-θ plane, the vertical line segment has a value in the vicinity of 0 °, so that an appropriate range, for example,
-5 ° <θ <5 °
Can be regarded as a vertical line segment. If n = 2, the points b and f are selected in the example of FIG. In addition, since the horizontal line segment has a value of θ around 90 °, an appropriate range, for example,
85 ° <θ <95 °
Can be regarded as a horizontal line segment. In the example of FIG. 8, the points a and e are selected. Needless to say, the range of θ shown here is merely an example and may be changed as appropriate.

  The second criterion is to divide the first image 108 into two along a vertical line centered on the first image 108, and in each range, the vertical length has a length equal to or greater than a certain ratio with respect to the vertical length of the first image 108. In addition to selecting the line segment in the direction, the first image 108 is divided into two on the horizontal line passing through the center thereof, and in each range, a length of a certain ratio or more with respect to the horizontal length of the first image 108 is set. This is a standard for selecting a horizontal line segment. Here, as an example, the fixed ratio is 70%.

At this time, the left half of the first image 108 is
0 ≦ r <X / 2
Similarly, the right half is X / 2 ≦ r <X
And the vertical line segment is −5 ° <θ <5 °.
Therefore, it is only necessary to select a line segment that satisfies the length condition within each range. In the example of FIG. 8, the point b and the point f are selected.

The lower half of the first image is
0 ≦ r <Y / 2
Similarly, the upper half is Y / 2 ≦ r <Y
And the vertical line segment is 85 ° <θ <95 °
Therefore, it is only necessary to select a line segment that satisfies the length condition within each range. In the example of FIG. 8, the points a and e are selected.

  The third criterion divides the first image 108 into two parts along a vertical line passing through the center as in the second criterion, and each area has a length equal to or greater than a certain ratio with respect to the longest vertical line segment. The first image 108 is divided into two on a horizontal line passing through the center of the first image 108, and in each range, the horizontal direction has a length equal to or greater than a certain ratio with respect to the longest horizontal line segment. This is a criterion for selecting the line segment. Here, as an example, the fixed ratio is 50%.

As mentioned above, the left half of the first image 108 is
0 ≦ r <X / 2
Similarly, the right half is X / 2 ≦ r <X
And the vertical line segment is −5 ° <θ <5 °.
Therefore, it is only necessary to select a line segment that satisfies the length condition within each range. In the example of FIG. 8, the point b and the point f are each selected as the longest vertical line segment. The point d is not selected because it does not reach the length of 50% of the line segment indicated by the point b.

The lower half of the first image is
0 ≦ r <Y / 2
Similarly, the upper half is Y / 2 ≦ r <Y
And the vertical line segment is 85 ° <θ <95 °
Therefore, it is only necessary to select a line segment that satisfies the length condition within each range. In the example of FIG. 8, the point a and the point e are selected as the longest horizontal line segment, and the point c is a line segment having a length equal to or greater than 50% of the line segment indicated by the point e. To be selected.

  Any of the criteria described above may be used, or criteria other than these may be used. When the second reference or the third reference is used, at least the extracted vertical and horizontal line segments are located on both sides of the center of the first image 108, respectively.

For the extracted line segment, a line segment that is substantially parallel to each edge of the image is selected from the selected line segments based on the distance from each edge, and a feature amount that characterizes the image is derived. The data is transferred to the feature quantity deriving unit 20 that functions as a quantity deriving unit. The feature quantity deriving unit 20 selects each edge of the first image 108, that is, a line segment substantially parallel to each side, from the selected line segments based on the distance from each side. In the present embodiment, the line segment closest to each side is selected. That is, specifically, for the left side, in FIG.
-5 ° <θ <5 °
The point with the smallest r in the range is selected. For the right side, a point where r is closest to X in the same range is selected. Also, for the lower side,
85 ° <θ <95 °
The point having the smallest r in the range is selected, and the point in the same range where r is closest to Y is selected for the upper side.

  In addition, the reference | standard at the time of selecting a line segment substantially parallel with respect to each edge | side from the selected line segment is not limited to the thing nearest to each edge | side. For example, a reference distance from each side may be determined in advance, and a line segment closest to the reference distance may be selected. This is effective when a rough distance from each side of a line segment constituting a frame to be selected is known in advance. Of course, line segments may be selected based on other criteria. In the above description, “substantially parallel” means a relationship that can be regarded as substantially parallel when image distortion is ignored. Specifically, the line segment 2 is extended. When the angle of the formed angle is equal to or less than a predetermined value, the two may be regarded as substantially parallel.

  The result of this operation will be specifically described for each reference adopted by the above-described line segment selection unit 19.

  First, when the first reference is adopted and when the second reference is adopted, the points indicating the selected line segment are point a, point b, point e, and point f. Since each of these points indicates a line segment closest to each side of the first image 108, the point a, the point b, the point e, and the point f are selected as they are. See also FIG.

  When the third criterion is adopted, points indicating the selected line segment are point a, point b, point c, point e, and point f. The points indicating the line segments closest to the left side, the right side, and the lower side of the first image are the point b, the point f, and the point a, respectively. For the upper side, the point representing the closest line segment is the point c. In this case, the point c is selected.

  Then, the feature amount deriving unit 20 derives the feature amount of the first image 100 based on the selected line segment. As described above, the specific value of the feature amount is not limited, but information indicating the position of the selected line segment, specifically, the position on the r-θ plane of each line segment ( The value of (r, θ) is one of feature quantities. Alternatively, the position of the intersection where the straight lines obtained by extending the selected line segments intersect is also one of the feature values. FIG. 9 is a diagram illustrating the position of the intersection point 110 that is the feature amount of the first image 108 when the line segment selection unit 19 adopts the third reference. In the figure, straight lines obtained by extending line segments a, b, c, and f are indicated by broken lines. In the present embodiment, the position of the intersection 110 is used as a feature amount. Of course, other values may be used as feature quantities.

  The first feature quantity that is the feature quantity of the first image 100 derived by the feature quantity deriving unit 20 is transferred to and stored in the first feature quantity storage unit 6 as shown in FIG. .

  Further, the feature quantity derived by the feature quantity deriving unit 20 is transferred to and stored in the feature quantity storage unit 21 that functions as a feature quantity storage unit that stores the feature quantity obtained for one image.

  Subsequently, it is assumed that the second image 200 is received by the image receiving unit 2 of the image processing apparatus 1 following the first image 100. Here, the second image 200 is as shown in FIG. 10, and the contents thereof are almost the same as those of the first image 100 (see FIG. 3). In the second image 200, the frame 201, the title column 202, the approval column 203, the outer frame 204, and the graphic 205 as a whole are inclined with respect to the outer shape of the second image 200, and the position and size of the frame 201 are also the first. This is slightly different from the frame 101 in the image 100 of FIG. In addition, the graphic 205 is corrected from the graphic 105 in the first image 100.

  The second image 200 is also corrected by the image correction unit 12, and then the line segment extraction range based on the feature amount stored in the feature amount storage unit 21 by the line segment extraction range restriction unit 13. There will be restrictions. In this process, only the range in the vicinity of the line segment selected in the first image 100 by the feature quantity deriving unit 20 is finally subjected to line segment extraction by the line segment extracting unit 18, and the other ranges are extracted. Restrict not to be subject to More specifically, using the feature amount stored in the feature amount storage unit 21, in this embodiment, the position of the intersection 110 (see FIG. 9), a straight line obtained by connecting the intersections is obtained. This line segment forms a rectangle. If the feature amount is a value of (r, θ) indicating the position on the r-θ plane of the line segment, a straight line constituting the rectangle is immediately obtained. Then, the vicinity of the obtained straight line, for example, a range from the straight line to a certain distance is set as a line segment extraction target, and the other range is excluded from the line segment extraction target. The method of excluding the line segment extraction target is, for example, painting out the entire range not covered by the line segment extraction with white, or excluding such a range from the processing in the partial extraction unit 14 and the line segment extraction unit 18. Also good.

  FIG. 11 is a diagram illustrating regions that are not subject to line segment extraction. In the figure, a straight line obtained by connecting the intersection 110 obtained for the first image 100 on the second image 200 shown in FIG. 10 is indicated by a broken line. And the boundary of the range from a straight line to the fixed distance is a dashed-dotted line, and hatching was performed to the other range which is not a line segment extraction object. By limiting the area indicated by hatching in FIG. 5 as being not subject to line segment extraction, noise and necessary calculation amount in the subsequent processing can be reduced.

  Note that the restriction by the line segment extraction range restriction unit 13 is based on the premise that the first image 100 and the second image 200 use substantially the same frames 101 and 201. Accordingly, when the second image 200 uses a frame that is significantly different from the first image 100, the line segment extraction range restriction unit 13 does not perform the restriction. This selection may be made by the user instructing the image processing apparatus 1 as appropriate. The feature amount storage unit 21 does not necessarily store only the feature amount of the image related to the immediately preceding process. For example, the feature amount storage unit 21 stores a plurality of feature amounts, and the user selects a feature amount based on which feature amount. You may make it select whether to limit a part extraction range.

  The subsequent operation is the same as that described for the first image 100. As a result, the second feature quantity that is the feature quantity of the second image 200 derived by the feature quantity deriving unit 20 is transferred to the second feature quantity storage unit 7 as shown in FIG. Remembered.

  When feature quantities are stored in both the first feature quantity storage unit 6 and the second feature quantity storage unit 7, selection is made based on the size of the region surrounded by the line segment selected by the feature quantity deriving means. The linear mapping deriving unit 8 that functions as a linear mapping deriving unit for deriving a linear mapping that matches the feature amount of one image with the feature amount of the other image compares the first feature amount and the second feature amount. Then, a linear map that matches the feature amount of the other image with the image selected based on the size of the region surrounded by the line segment indicated by each feature amount is obtained. In the present embodiment, as an example of selection criteria, an image having a large area surrounded by the line segment indicated by the feature value is selected. The linear mapping obtained by the linear mapping deriving unit 8 is the position of the intersection of the straight lines that form the rectangle, and therefore the rectangle formed by each intersection is large. The linear map is such that the other intersection coincides with one of the other intersections. There are various types of such linear mapping. Here, a matrix which is a transformation coefficient of the linear mapping called affine transformation is obtained. The affine transformation is a transformation obtained by synthesizing translation, rotation, and enlargement / reduction in a two-dimensional plane. Note that a known method may be used as a method for obtaining a conversion coefficient of affine transformation, and detailed description thereof is omitted in this specification.

  In the present embodiment, the linear mapping deriving unit 8 determines the linear mapping for the two images, but is not limited to this, and may determine the linear mapping for three or more images. Even in this case, it is preferable to obtain a linear map that matches the feature amount of another image with the image selected based on the size of the region surrounded by the line segment indicated by the feature amount of each image.

  Further, the image conversion unit 9 uses the conversion coefficient obtained by the linear mapping derivation unit 8 to use the first image obtained from the first image storage unit 3 or the second image obtained from the second image storage unit 4. One of the images is converted, and both are transferred to the image comparison unit 10.

  The image comparison unit 10 compares the two received images and generates an image that can identify the place where the change has occurred. More specifically, the difference between the two transferred images is taken, and the obtained image is transferred to the difference image output unit 11 as a difference image. Of course, the comparison method is not limited to taking the difference, and other methods may be used.

  The difference image output unit 11 outputs the delivered difference image. The output format is not particularly limited, and it may be transmitted to other information terminals through a telecommunication line, displayed on a monitor, printed by a printer, or recorded on an arbitrary information recording medium. Also good.

  Here, the first image 100 shown in FIG. 3 and the second image 200 shown in FIG. 10 are transferred to the image processing apparatus 1, and the linear mapping deriving unit 8 sets conversion coefficients for converting the second image 200. The case where it derives is assumed as an example.

  In this case, the second image 200 is converted by the image conversion unit 9, and a second image 206 whose inclination, position, and size are corrected as shown in FIG. 12 is obtained.

  Then, the difference between the first image 100 and the second image 206 is taken by the image comparison unit 10, and a difference image 300 shown in FIG. 13 is obtained. The difference image 300 is an image illustrating only a portion that is different between the first image 100 and the second image 206.

  As described above, the image processing apparatus 1 according to the present embodiment arranges line segments constituting a frame included in an image even when a plurality of images are deformed by rotation, parallel movement, enlargement / reduction, and the like. Is calculated correctly, and a feature amount is derived based on the arrangement of the line segment. Then, since the deformation between the plurality of images is removed by the linear mapping obtained based on the feature amount derived in this way, even if the degree of the deformation is larger than a certain limit, one Will be compared correctly with other images.

  In the above description of the present embodiment, the image processing apparatus 1 compares two images. However, the present invention is not limited to this, and three or more images may be compared with each other. In that case, another image may be compared with the designated image.

  In the above description, specific functional block configurations of the image processing apparatus 1, the linear mapping derivation unit 8, and the image processing apparatus 1 are shown as an example, and the present invention is not limited thereto. The technical scope of the present invention is defined based on the description of the scope of claims.

  DESCRIPTION OF SYMBOLS 1 Image processing apparatus, 2 Image reception part, 3 1st image storage part, 4 2nd image storage part, 5 Image feature-value derivation | leading-out part, 6 1st feature-value storage part, 7 2nd feature-value storage part 8 Linear mapping derivation unit 9 Image conversion unit 10 Image comparison unit 11 Difference image output unit 12 Image correction unit 13 Line segment extraction range restriction unit 14 Partial extraction unit 15 Contour figure acquisition unit 16 Connected figure Identification unit, 17 Maximum connected figure acquisition unit, 18 line segment extraction unit, 19 line segment selection unit, 20 feature amount derivation unit, 21 feature amount storage unit, 100 first image, 101 frame, 102 title column, 103 approval column 104 Outer frame, 105 figure, 106, 107, 108, 109 First image, 110 Intersection, 200 Second image, 201 frame, 202 Title field, 203 Approval field, 204 Outer frame, 205 Figure, 206 Second image, 300 difference image.

Claims (6)

A line segment extracting means for extracting a line segment included in the image;
Of the extracted line segments, line segment sorting means for sorting line segments based on a predetermined criterion;
Feature amount deriving means for deriving a feature amount characterizing the image by selecting a line segment substantially parallel to each edge of the image from the selected line segments based on a distance from the edge; ,
Linear mapping derivation means for deriving a linear mapping that matches the feature quantity of another image with the feature quantity of one image selected based on the size of the region surrounded by the line segment selected by the feature quantity derivation means;
Image comparing means for comparing the one image with another image transformed by the linear mapping derived by the linear mapping deriving means;
An image processing apparatus.   The image processing apparatus according to claim 1, further comprising a partial extraction unit that extracts a part of a graphic included in the image and newly sets the part as the image.   3. The image according to claim 2, wherein the partial extraction unit includes a contour graphic acquisition unit that acquires a contour graphic that is a graphic indicating a contour of a graphic obtained by a painting process starting from at least one point of the outer edge or the vicinity of the outer edge of the image. Processing equipment. The partial extraction means includes
Connected figure identifying means for identifying a connected figure that is a connected figure included in the image;
Among the connected figures, maximum connected figure acquisition means for acquiring the maximum connected figure which is the maximum connected figure,
The image processing apparatus according to claim 2, further comprising: Feature quantity storage means for storing the feature quantity obtained for one of the images;
A line segment extraction range limiting unit that limits a range in which the line segment extraction unit extracts a line segment for the other image based on the feature amount stored in the feature amount storage unit;
An image processing apparatus according to claim 1, comprising: Computer
A line segment extracting means for extracting a line segment included in the image;
Of the extracted line segments, line segment sorting means for sorting line segments based on a predetermined criterion;
Of the selected line segments, a line segment that is substantially parallel to each edge of the image is selected based on the distance from each edge, and a feature value deriving unit that derives a feature value that characterizes the image;
Linear mapping derivation means for deriving a linear mapping that matches the feature quantity of another image with the feature quantity of one image selected based on the size of the region surrounded by the line segment selected by the feature quantity derivation means;
Image comparing means for comparing the one image with another image transformed by the linear mapping derived by the linear mapping deriving means;
A computer program that functions as a computer program.

Images