JPH0145103B2 - - Google Patents

Description

[Detailed Description of the Invention] (a) Technical Field of the Invention The present invention relates to an image processing device, which processes two congruent or similar images on the same plane and processes changes in pixel values for each pixel of each image. The present invention relates to an image-to-image rotation angle difference verification device that verifies whether or not a rotation angle difference between both images determined from an angle histogram regarding directional angles is correct.

(b) Background of the technology In a seal matching device, which is an applied field of image processing, usually the seal impression to be compared, that is, the image to be matched, and the registered seal impression, that is, a reference image, stored in advance in a dictionary, etc. are used. , a process of matching on the same plane using a technique such as pattern matching to find the degree of similarity is often used, but the direction (orientation) of the matched seal impression and the registered seal impression are different, that is, the matched seal impression and the registered seal impression. If there is a difference in rotation angle between the

Therefore, in such cases, in the field of image processing or image recognition, it is necessary to perform processing to remove the rotation angle difference mentioned above in the preprocessing prior to matching, and to remove the difference between two congruent or similar images. The process of determining the rotation angle difference between the two is often performed.

(c) Prior art and problems In order to find the rotation angle difference between multiple images that are congruent or similar on the same plane, conventionally the image is divided into pixels arranged in a mesh pattern and the pixel value of each pixel is calculated. An angle histogram generation unit that generates an angle histogram by calculating the angle in the direction in which the angle changes from large to small or from small to large for each pixel and counting the number of pixels for each angle. The difference in rotation angle between two images is determined by comparing the angle histograms obtained by different parts of the image.

However, since characters or line figures are generally composed of nearly parallel lines, for example, in a line figure such as the one shown in Figure 1a, the direction in which the density value changes at each pixel is indicated by the arrow. Therefore, an angle histogram as shown in b is obtained.

Similarly, for a figure such as c, which is the figure in a rotated by 180 degrees, an angle histogram d that is very similar to b is obtained, and in this way, the rotation angle difference between the two images can be expressed as an angle histogram. When calculating the angle, a 90 degree error is likely to occur, so there was a problem in that the detection results had to be verified. Also, since the lines that make up characters often have orthogonal components,
There was a problem that depending on the characters, errors of 90 degrees or 270 degrees were likely to occur.

In FIG. 2, the image c is obtained by rotating the image a by 45 degrees, b is the angle histogram for the image a, and d is the angle histogram for the image c. By comparing the angle histogram shown by b and the angle histogram shown by d while moving them relatively in the angular axis direction, the image shown by a and the image shown by c are compared. The rotation angle difference between the image and the displayed image is -45 degrees/45
It turns out that it is 135 degrees or 225 degrees.

To verify this, we need to create an image obtained by rotating the image c by 45 degrees, an image obtained by rotating it by -45 degrees, and an image obtained by rotating it by -135 degrees. All you have to do is compare the image rotated by 135 degrees with the image a, check the degree of similarity (distance), and select the one with the highest degree of similarity, but image rotation takes a long time. The problem was that it had to be rotated four times in this way.

(d) Object of the Invention The object of the invention is to obtain two images that are congruent or similar on the same plane and that are obtained from an angular histogram regarding the angle in the direction of change in pixel value for each pixel of each image. The object of the present invention is to provide a means for quickly verifying whether the rotation angle between the two is correct or not.

(e) Structure of the Invention In other words, the inter-image rotation angle verification device of the present invention divides an image into pixels arranged in a mesh pattern and calculates the angle in the direction in which the pixel value of each pixel changes from large to small or from small to large. an angle histogram generation unit that calculates the number of pixels for each pixel and counts the number of pixels for each angle to generate an angle histogram; and a comparison unit that compares two angle histograms obtained by the angle histogram generation unit for two images. and an image rotation used in an inter-image rotation angle difference verification device that detects a rotation angle difference between the two images, and that rotates at least one of the two images based on the detected rotation angle difference. a projection extraction unit that extracts projections on two mutually intersecting axes from each of the two images from which the rotation angle difference has been removed by the image rotation unit; and a distance calculation section that calculates a distance between projections of images, and the detected rotation angle difference is verified by the projection obtained by the projection extraction section.

(f) Examples of the Invention Next, the gist of the present invention will be specifically explained using examples.

Figures 3 and 4 are diagrams for explaining the principle of the present invention.
The images obtained by applying rotations of 135 degrees and -135 degrees, and Figure 4 a, b, c and d are x of the images represented by Figure 3 a, b, c and d, respectively. It shows the projection onto the coordinate axis (horizontal axis) and the y-coordinate axis (vertical axis).

Comparing each projection in Fig. 4, we find that x of a
The coordinate axis projection, the y-coordinate axis projection of b, the inversion of the x-coordinate axis projection of c in the x-coordinate axis direction, and the inversion of the y-coordinate axis projection of d in the y-coordinate axis direction are equal, and the y-coordinate axis projection of a, It can be seen that the x-coordinate axis projection of b is inverted in the x-coordinate axis direction, the y-coordinate axis projection of c is inverted in the y-coordinate axis direction, and the x-coordinate axis projection of d is equal.

In other words, as shown in Figure 3 a, b, c and d, 90
For images with rotation angle differences of degrees, it is difficult to verify the rotation angle difference using an angle histogram, but it can be easily verified by using projections on the x and y axes, respectively. , and if the projection of an image to two orthogonal axes is obtained, the image can be projected at 90 degrees, 180 degrees, and 270 degrees.
It can be seen that projections of the rotated image on each of the two axes can be obtained very easily.

The present invention has been made based on the above principle, and FIG. 5 is a block diagram showing one embodiment thereof.

In Fig. 5, 1 is an image memory that divides the image of the registered seal imprint into pixels arranged in a mesh pattern and stores the pixel value of each pixel, and 2 divides the image of the seal impression to be verified into pixels arranged in a mesh pattern. The image memory 3 stores the pixel value of each pixel, and 3 calculates the angle in the direction in which the pixel value of each pixel of each image stored in the image memory 1 and the image memory 2 changes from large to small for each pixel. An angle histogram generation unit that counts the number of pixels to generate an angle histogram; and a comparison unit that compares two angle histograms obtained by the angle histogram generation unit for the two images. An angular difference detection device 4 detects a rotation angle difference between the images;
6x is a projection extraction unit that extracts the x-coordinate axis projection of the image stored in the image memory 1; and 6y is the y-coordinate axis of the image stored in the image memory 1. A projection extraction unit 7x extracts a projection, and 7x is a projection extraction unit 7 that extracts an x-coordinate axis projection of an image stored in the image memory 5.
y is a projection extraction unit that extracts the y-coordinate axis projection of the image stored in the image memory 5, 8x is a memory that stores the projection extracted by the projection extraction unit 6x, and 8y is the projection extracted by the projection extraction unit 6y. A memory for storing projections; 9x is a memory for storing projections extracted by the projection extraction unit 7x; 9y is a projection extraction unit 7;
a memory storing the projection extracted by y, 1
0 is a distance calculation unit that calculates and compares the distances between the projections stored in the memory 8x, the memory 8y, the memory 9x, and the memory 9y, and the projections obtained by inverting these projections in each coordinate axis direction.

FIG. 6 is a block diagram showing a detailed configuration example of the angular difference detection device 3, in which the same reference numerals as in FIG. 5 refer to the same objects. Buffers for pipeline processing that sequentially store a pixel value and the pixel values of three pixels in the next row, 3a, 4a, 5a, and 6a are used to store the pixel values of the image pixels stored in the image memory 1 from large to small. An angle histogram generation unit 3A is constructed which calculates the angle in the direction of change for each pixel and counts the number of pixels for each angle to generate an angle histogram.
4a is a difference calculation circuit that reads the pixel value of each pixel and calculates the difference dx in the row direction (x direction);
Read the pixel value of each pixel of 3 and move it in the column direction (y
Difference calculation circuit for calculating the difference dy in direction), 5a
is the angle arctan(dy/dx) in the direction in which the pixel value changes from large to small from the difference dy obtained by the difference calculation circuit 4a and the difference dx obtained by the difference calculation circuit 3a for each pixel. An angle calculation circuit 6a stores the histogram of the output of the angle calculation circuit 5a. 2b reads the contents of the image memory 2 and calculates the pixel values of the pixels in two rows and the three pixels in the next row. Buffers 3b, 4b, 5b, and 6b for pipeline processing that sequentially store values and values calculate the angle in the direction in which the pixel value of the pixel of the image stored in the image memory 2 changes from large to small for each pixel. An angle histogram generation unit 3B that counts the number of pixels at each time and generates an angle histogram is configured, and 3b is a buffer 2.
4b is a difference calculation circuit that reads the pixel value of each 3×3 pixel shown by hatching and calculates the difference dx in the row direction (x direction); A difference calculation circuit 5b reads the pixel value of each pixel of 3 and calculates the difference dy in the column direction (y direction). 6b is a memory for storing the output histogram of the angle calculation circuit 5b; 11 is a memory This is a comparison section that compares the angular histograms stored in the angular histogram 6a and the memory 6b while shifting them relatively in the respective angular axis directions.

To explain the case where images such as those illustrated in FIGS.
2, the angular histograms as shown in FIG.
and is stored in the memory 6b, and the comparator 11 outputs 45 degrees as the rotation angle difference.

The image rotation unit 4 rotates the image stored in the image memory 2 by 45 degrees clockwise, and as a result, a figure like a is stored in the image memory 5.

The projection extraction unit 6x calculates x by counting the number of pixels in each pixel column whose pixel value is equal to or higher than a predetermined threshold among the pixels constituting the image stored in the image memory 1.
The projection extraction unit 6y extracts the coordinate axis projection, and the projection extraction unit 6y calculates the y coordinate axis projection by counting the number of pixels in each pixel row whose pixel value is greater than or equal to a predetermined threshold among the pixels constituting the image stored in the image memory 1. The extracted projections are stored in memory 8x and memory 8y, respectively.
is stored in

Similarly, the projection extraction unit 7x extracts the x-coordinate axis projection by counting the number of pixels in each pixel column whose pixel value is equal to or higher than a predetermined threshold among the pixels constituting the image stored in the image memory 5. Then, the projection extractor 7
y is the y-coordinate axis projection extracted by counting the number of pixels in each pixel row whose pixel value is greater than or equal to a predetermined threshold among the pixels constituting the image stored in the image memory 5, and each extracted projection is It is stored in memory 9x and memory 9y.

The distance calculation unit 10 first reads the memory 8 for the first time.
The sum D1 of the distance between the projection stored in x and the projection stored in the memory 9x and the distance between the projection stored in the memory 8y and the projection stored in the memory 9y is determined.

Next, for the second time, the distance between the projection stored in the memory 8x and the projection obtained by reversing the projection stored in the memory 9y in the y-coordinate axis direction, and the distance between the projection stored in the memory 8y and the projection stored in the memory 9y are calculated. The sum D2 of the distance between 9x and the stored projection is calculated.

Subsequently, in the third time, the distance between the projection stored in the memory 8x and the projection obtained by reversing the projection stored in the memory 9x in the x coordinate axis direction, and the projection stored in the memory 8y and the memory are calculated. The sum D3 of the distance between the projection stored in 9y and the projection obtained by inverting it in the y-coordinate axis direction is calculated.

Finally, the distance between the projection stored in the memory 8x and the projection stored in the memory 9y is obtained by reversing the projection stored in the memory 8y and the projection stored in the memory 9x in the y-coordinate axis direction. The sum D4 of the distance between the two projections is calculated.

That is, in the first time, the distance D1 between the projection extracted from the image in FIG. 2e and the projection extracted from the image in FIG.
The distance D2 between the projection extracted from the image in figure a and the projection that would have been extracted from the image in figure 3 b is obtained, and in the third time the projection extracted from the image in figure 2 a and The distance D3 between the projection and the one that should be extracted from the image in Figure 3c is obtained, and the second
The distance D4 between the projection extracted from the image of figure a and the projection to be extracted from the image of figure 3d is obtained.

That is, by comparing D1, D2, D3, and D4, the rotation angle difference between the image stored in image memory 1 and the image stored in image memory 2 is determined as follows.
If D1 is the minimum, it is the value detected by the angular difference detection device 3, and if D2 is the minimum, it is the value detected by the angular difference detection device 3 plus 90 degrees, If D3 is the minimum, the angle difference detection device 3
is the value found by adding 180 degrees,
Furthermore, if D4 is the minimum, it can be verified that it is a value obtained by adding 270 degrees to the value detected by the angular difference detection device 3.

According to the above embodiment, the image rotation unit 4 only needs to rotate the image of the seal imprint to be verified stored in the image memory 2 once.

Further, in the distance calculating section 10, for one image stored in the image memory 5, for example, FIG.
Once the x-coordinate axis projection and y-coordinate axis projection as shown in figure a are obtained, the image in figure 3 a can be rotated by 90 degrees, 180 degrees and 270 degrees as shown in figures b, c and d, respectively. The x-coordinate axis projection and y-coordinate axis projection for the image obtained can be obtained extremely easily by simply changing the order in which data is read from the memory 9x or the memory 9y, respectively.

Therefore, the rotation angle difference obtained by the image rotation section 3 can be verified easily in a short time.

(g) Effects of the Invention As explained above, according to the present invention, for two congruent or similar images on the same plane, it is possible to calculate the angle histogram of the change direction of the pixel value of each pixel of each image. It is possible to verify in a short time whether the obtained rotation angle between both images is correct.

[Brief explanation of drawings]

1 and 2 are explanatory diagrams of problems in the conventional example, FIGS. 3 and 4 are explanatory diagrams of the principle of the present invention, FIG. 5 is a configuration diagram of an embodiment of the present invention, and FIG. A detailed configuration diagram of the angular difference detection device is shown. In the figure, 3 is an angle difference detection device, 4 is an image rotation unit,
6x, 6y, 7x and 7y are projection extractors, 10
3a, 4a, 4b and 3b are difference calculation circuits; 5a and 5b are angle calculation circuits; 1
1 is a comparison section.

Claims (1)

[Claims] 1 Divide the image into pixels arranged in a mesh pattern, find the angle in the direction in which the pixel value of each pixel changes from large to small or from small to large for each pixel, count the number of pixels for each angle, and create an angle histogram. and a comparison unit that compares two angle histograms obtained by the angle histogram generation unit for two images, and a rotation angle that detects a rotation angle difference between the two images. an image rotation unit that is used in a difference detection device and rotates at least one of the two images based on the detected rotation angle difference; and two images from which the rotation angle difference has been removed by the image rotation unit. a projection extraction unit that extracts projections from each of the two axes that intersect with each other, and a distance calculation unit that calculates a distance between the projections of the two images extracted by the projection extraction unit, An image-to-image rotation angle difference verification apparatus, characterized in that the rotation angle difference between images is verified using a projection obtained by the projection extraction section.