JPH0981729A - Image matching method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform a matching process with high precision by canceling the deformation of an object image to be matched even if the object image is deformed. SOLUTION: The object image 2 to be matched is matched against matching reference images 1a-1d while shifted in position to obtain the position where the peak degree of matching is obtained and the peak degree of matching and obtain matching positions α, β, γ, and δ including the degrees Ma-Md of matching of the object image 2 against the matching reference images 1a-1d and the deviations, and the deviations of the matching positions are corrected to calculate a matching position P and a matching degree M, including no deviation, which corresponds to the matching position and matching degree obtained when a pattern in the same shape with the matching object image 2 is decided as a matching reference image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method, and more particularly to an image matching method for correcting a systematic shift or the like caused by deformation of a matching target image.

[0002]

2. Description of the Related Art In a conventional image matching device,
Many of them simply apply normal correlation. As an example of a conventional technique that has improved it, Japanese Patent Laid-Open No. Sho 60-20037.
The invention of the template matching method described in Japanese Patent Publication No. 5 (hereinafter, referred to as Publication 1), and Japanese Patent Laid-Open No.
There is an invention of a pattern matching method described in Japanese Patent Publication No. 156383 (hereinafter, referred to as Publication 2).

In the invention of the template matching method described in the publication 1, the template is weighted in a predetermined manner so that it exists at the boundary between the recognized pattern and the background pattern in the input digital image pattern. The purpose is to improve the detection accuracy by eliminating the influence of unstable elements such as deformation.

Further, the invention of the pattern matching method described in the publication 2 is multi-valued by performing blur processing by averaging two image patterns of a dictionary pattern and a target pattern in a local region. 2 as a pattern
The object of the present invention is to perform template matching of two image patterns so that the overlapping portion can be enlarged and recognition can be performed even if the positions of the two image patterns are slightly displaced.

[0005]

First, in the conventional image matching apparatus, when the matching target image is deformed with respect to the matching reference image, the peak matching degree decreases, and at the same time, the matching degree peaks. There is a problem that a deviation occurs in the matching position.
Further, even in an image matching device that takes improvement measures, the matching degree is further reduced due to the deformation of the matching target image, and a statistical error occurs in the peak matching position, so that the calculation accuracy of the peak matching position is improved. There is a problem that it is difficult to secure.

[0006] These reasons are based on the assumption that there is no deformation other than the random noise component between the matching target image and the matching reference image when performing the correlation matching, and in the case of random noise. This means that the offset that is statistically canceled remains statistically in the deformation of the matching target image, and the statistical offset of the matching position of the remaining peak is not considered.

Secondly, particularly when matching is performed at a corner, if the corner of the target image is directly matched with the corner reference image when the corner angle varies, the accuracy of the corner position deteriorates. .
Further, when matching two sides of the corner with the edge reference image, it is necessary to perform matching processing with the edge reference image very finely by changing the angle in order to ensure the accuracy of calculating the intersection of the corners. There is.

[0008] The reason is that, especially when performing corner matching processing, the corner area specified by the matching reference image is matched, not the point at the corner tip position is matched. That is, since the two edges of the corner image are substantially targeted for matching, the occurrence of offset is large.

In view of such a point, the first object of the present invention is to cancel the fluctuation of the matching target image and perform the highly accurate matching process. The second purpose is to estimate the amount of the variation as needed. A third object is to improve the matching positioning accuracy when the corner angle changes.

[0010]

In order to solve such a problem, an image matching method of the present invention has a plurality of average patterns or representative patterns as a plurality of matching reference images, and the plurality of matching reference images are provided. Matching is performed while shifting the position between each of the images and the matching target image, and as a result of performing the matching, each of the plurality of matching reference images has a position at which a matching degree peak is obtained and the peak. Of the matching reference image and the matching target image, and a deviation caused by deformation or mismatch between the matching reference image and the matching target image. And a plurality of matching positions including the deviation, and a plurality of matching positions including the deviation. To correct the deviation of the matching position of the plurality of the, equivalent to that obtained in the case of the pattern having the same shape as the matching target image as a matching reference image, and calculates the matching position and the matching degree without the deviation.

In the image matching method of the present invention, the matching as a method for obtaining the position where the peak of the matching degree is obtained and the matching degree of the peak is combined with pixel-based matching and subpixel interpolation. ,
The plurality of matching positions are determined as the estimated peak matching degree obtained by performing the sub-pixel interpolation on the matching degree directly obtained in the vicinity of the peak position of the matching degree directly obtained by the matching. Can be a matching position obtained by performing the sub-pixel interpolation corresponding to the estimated peak matching degree.

In the image matching method of the present invention, the matching may be normal correlation or simple correlation.
With the plurality of matching positions as correlation positions obtained by performing the normal correlation or correlation positions obtained by performing the simple correlation, the plurality of matching degrees are obtained as correlation values obtained by performing the normal correlation or the simple correlation. It can be a correlation substitute value obtained by performing the calculation.

The image matching method of the present invention has a plurality of average patterns or representative patterns as a plurality of matching reference images, and shifts the position between each of the plurality of matching reference images and the matching target image. Matching is performed, and as a result of the matching, a position at which a peak of the matching degree is obtained for each of the plurality of matching reference images and the matching degree of the peak are obtained, and the plurality of matching reference images are obtained. A plurality of matching degrees between the matching target image and a plurality of matching positions including a deviation caused by deformation or mismatch between the plurality of matching reference images and the matching target image are obtained, and the deviation is included. Correcting deviations of the plurality of matching positions from the plurality of matching positions and the plurality of matching degrees, Equivalent to that obtained when the pattern of the same shape as the matching target image and the matching reference image, the amount of deformation of the matching target image, calculating at least one of the dimensions or angles.

In the image matching method of the present invention, the plurality of matching reference images have a blurred corner image obtained by averaging images near an average corner angle or a non-blurred corner image having a typical corner angle. When,
A corner image with blur that is obtained by averaging images near the maximum corner angle or a non-blurring corner image that has the maximum corner angle, and a corner image or minimum corner that has blur that is obtained by averaging images near the minimum corner angle. It is possible to have three images with an unblurred corner image with an angle.

Further, in the image matching method of the present invention, the matching as a method of obtaining the position at which the peak of the matching degree is obtained and the matching degree of the peak, pixel-based matching and sub-pixel interpolation are performed. The plurality of matching degrees are used together as the estimated peak matching degree obtained by performing the sub-pixel interpolation on the matching degree directly obtained in the vicinity of the peak position of the matching degree directly obtained by the matching. Can be the matching position obtained by performing the sub-pixel interpolation corresponding to the estimated peak matching degree.

According to the image matching method of the present invention, a circle image having blur on the edge portion obtained by averaging circle images whose centers coincide with each other in the vicinity of an average size or a circle image having no blur of a typical size, and a minimum A circle image with blur on the edge part obtained by averaging circle images with matching centers near the size or a circle image with minimum blur, and a circle image with matching centers near the maximum size, obtained by averaging A plurality of matching reference images, each of which has a circle image having blur on the edge portion thereof or a circle image having no blur at the maximum size, and which is positioned between each of the plurality of matching reference images and the matching target image. The matching is performed while shifting, and as a result of the matching, each of the plurality of matching reference images has a degree of matching degree peak. And a degree of matching of the peak, and a plurality of degree of matching between the plurality of matching reference images and the matching target image, a transformation between the plurality of matching reference images and the matching target image, or A plurality of matching positions including deviations caused by mismatches are obtained, and deviations of the plurality of matching positions are corrected from the plurality of matching positions including the deviations and the plurality of matching degrees to obtain the same shape as the matching target image. At least one of the deformation amount, the size, and the position of the matching target image, which corresponds to the pattern obtained when the pattern is used as the matching reference image, is calculated.

[0017]

BEST MODE FOR CARRYING OUT THE INVENTION As a matching reference image, a plurality of images such as an image having an average or typical characteristic, an image having a maximum characteristic, an image having a minimum characteristic are used. Matching processing is performed between the plurality of matching reference images and the matching target image to be recognized, and the peak matching position and the matching degree are obtained. Interpolation processing is performed using the obtained matching position and matching degree, and the matching position and matching degree that would be obtained when performing peak search with a matching reference image having the same characteristics as the matching target image are estimated.

In the case of a corner image, three images of a corner image having an average or typical corner angle, a corner image having a maximum corner angle, and a corner image having a minimum corner angle are used as matching reference images. use. At this time, the vertex position and the corner direction of each reference image are matched. A matching search for finding the peak of the matching degree is performed between the three matching reference images and the matching target image while shifting the position of the matching reference image in pixel units, and the matching degree of the peak and the matching position corresponding thereto are obtained. Further, if necessary, the matching degree around the peak position of the matching degree is interpolated to obtain a matching position corresponding to the peak matching degree in the sub-pixel. In this way, the interpolation processing is performed using the matching degrees of the plurality of peaks obtained for the plurality of matching reference images and the matching positions corresponding thereto, and the interpolation processing is performed from the corner angle and the corner position of the matching reference image. Then, the deformation amount is calculated, and the corner angle and the corner position of the matching target image are calculated.

In the case of a circle image, three images of an average or representative size circle image, a minimum size circle image, and a maximum size circle image are used as matching reference images. A matching search for finding the peak of the matching degree is performed between the three matching reference images and the matching target image while shifting the position of the matching reference image in pixel units, and the matching degree of the peak and the matching position corresponding thereto are obtained. Further, if necessary, the matching around the peak position of the matching degree is interpolated to obtain the matching position corresponding to the matching degree of the peak in the sub-pixel. In this way, interpolation processing is performed using the matching degrees of the plurality of peaks obtained for the plurality of matching reference images and the corresponding matching positions, and the interpolation processing is performed from the radius and center position of the matching reference images. The amount of deformation is calculated, and further, the radius and center position of the matching target image are calculated.

[0020]

Next, embodiments of the present invention will be described with reference to the drawings.

[First Embodiment] FIG. 1 is a diagram showing a process in a first embodiment of the present invention. By using correlation as a matching method, the matching position and the degree of matching are correlated position and correlation value. I am trying. FIG. 1A is a diagram showing the configuration, and FIG.
FIG. 6 is a diagram showing a relationship between a correlation position and a correlation value.

FIG. 1A shows a structure having a plurality of matching reference images 1, a matching target image 2, a correlation peak search mechanism 3 and an interpolation system 4.

The plurality of matching reference images 1 are images stored in advance, and three types of images, that is, an image having an average characteristic, an image having a maximum characteristic, and an image having a minimum characteristic, are the lowest. I only need it. In the first embodiment, four types of 1a, 1b, 1c and 1d are used.

The matching target image 2 is an image to be recognized and includes a plurality of matching reference images 1a, 1b, 1
Matches with c and 1d.

The correlation peak search mechanism 3 is an algorithm for finding the peak position of a wide range of correlation values on a pixel-by-pixel basis. Subpixel interpolation is applied near the peak position. The correlation peak search mechanism 3 correlates the correlation positions α, β, γ, δ and the correlation value M between the plurality of matching reference images 1a, 1b, 1c, 1d and the matching target image 2.
a, Mb, Mc, Md are calculated.

The interpolating system 4 includes the correlation positions α, β, γ, δ and the correlation value M obtained by the correlation peak search mechanism 3.
Interpolation processing of a, Mb, Mc, and Md is performed to obtain the final correlation position P and correlation value M.

As shown in FIG. 1B, the final correlation position P and the correlation value M thus obtained are the correlation position and the correlation value obtained from the matching reference image having the same characteristics as the matching target image 2. Interpolate and estimate the correlation value.

In the first embodiment, as the matching reference image, an image having an average characteristic, an image having the maximum characteristic, and an image having the minimum characteristic are used.
Although the type is used, another image such as an image having typical characteristics may be used. Also, as a matching method, sub-pixel interpolation was used together with pixel-by-pixel correlation.
As a method of this correlation, methods such as normal correlation and simple correlation can be used. Furthermore, subpixel interpolation may not be used together.

[Second Embodiment] FIG. 2 is a diagram showing a process in the second embodiment of the present invention. By using correlation as a matching method, the matching position and the degree of matching are determined as the correlation position and the correlation value. I am trying. FIG. 2A is a diagram showing the configuration, and FIG.
FIG. 6 is a diagram showing a relationship between a correlation position and a correlation value. In FIG. 2, the corner angle of the matching target image is obtained by the interpolation processing from the corner angle serving as the reference of the matching reference image and the deformation amount (displacement angle) obtained by the interpolation processing, and becomes the reference of the matching reference image. The case where the corner position of the matching target image is obtained from the corner position and the deformation amount (position displacement amount) obtained by the interpolation processing is shown.

FIG. 2 shows a corner image 5 having an average corner angle b and a corner image 6 having a maximum corner angle c.
And a corner image 7 having a minimum corner angle a, a matching target image 8, and an image (hereinafter, referred to as an image) 9 obtained by matching.

Three images, a corner image 5 having an average corner angle b, a corner image 6 having a maximum corner angle c, and a corner image 7 having a minimum corner angle a, are used as matching reference images. At this time, the vertex positions and the corner directions of the matching reference images 5, 6, and 7 are matched.

The matching target image 8 is matched with each of the matching reference images 5, 6 and 7 to obtain an image 9.

In the image 9, the correlation positions obtained for the matching reference images 7, 5, 6 are α, β, γ, respectively, and are obtained for the matching reference images 7, 5, 6. Correlation values are Ma, Mb, and Mc, respectively.
It is.

As shown in FIG. 2 (b), the correlation positions α, β, γ and the correlation values Ma, M thus obtained are obtained.
The corner angle d that maximizes the correlation value M and the corner position δ at the corner angle d, that is, the matching target image 8 is provided as indicated by 8 * in FIG. The corner position δ at the corner angle d is calculated.

In the second embodiment, a corner image having an average corner angle is used as a matching reference image,
Although three types of images, a corner image having the maximum corner angle and a corner image having the minimum corner angle, are used, other images such as a corner image having a typical corner angle may be used. Further, as a matching method, subpixel interpolation may be used in combination with pixel-by-pixel correlation, and a normal correlation method, a simple correlation method, or the like may be used as a correlation method.

[Third Embodiment] FIG. 3 is a diagram showing a process in a third embodiment of the present invention. By using correlation as a matching method, the matching position and the matching degree are determined as the correlation position and the correlation value. I am trying. FIG. 3A is a diagram showing the configuration, and FIG.
FIG. 6 is a diagram showing a relationship between a correlation position and a correlation value. In FIG. 3, by the interpolation processing, from the reference radius of the matching reference image and the deformation amount obtained by the interpolation processing,
The figure shows a case where the circle size of the matching target image is obtained and the central position of the matching target image is obtained from the center position serving as the reference of the matching reference image and the deformation amount obtained by the interpolation processing.

FIG. 3 shows a circular image 1 with an average radius e.
0, a circular image 11 with a maximum radius f, and a minimum radius g
A circular image 12 including a matching target image 13,
An image (hereinafter referred to as an image) 14 obtained by matching is configured.

Three images, a circular image 10 having an average radius e, a circular image 11 having a maximum radius f, and a circular image 12 having a minimum radius g, are used as matching reference images. At this time, each matching reference image 10, 1
The center positions of 1 and 12 are matched.

The matching target image 13 is matched with each of the matching reference images 10, 11 and 12 to obtain an image 14.

In the image 14, the correlation radii obtained for the matching reference images 12, 10, 11 are ε, ζ, η, respectively.
The correlation values obtained for 0 and 11 are Mg and
Me and Mf.

As shown in FIG. 3 (b), the correlation radii ε, ζ, η and the correlation values Mg, M thus obtained are obtained.
e and Mf are interpolated, and the radius h that maximizes the correlation value M and the center position θ at the radius h, that is, the radius h included in the matching target image 13 as indicated by 13 * in FIG. The center position θ at is calculated.

In the third embodiment, three types of matching reference images, a circular image having an average radius, a circular image having a maximum radius, and a circular image having a minimum radius, are used. Other images may be used, such as a circular image with a representative radius. Further, as a matching method, subpixel interpolation may be used in combination with pixel-by-pixel correlation, and a normal correlation method, a simple correlation method, or the like may be used as a correlation method.

[0043]

As described above, the present invention has the following effects.

First, there is a deformation in the matching target image by performing an interpolation process using a plurality of matching reference images, estimating the influence of the deformation of the matching target image, and calculating the matching position and the matching degree. Even in such a case, there is an effect that the fluctuation can be canceled and the matching process can be performed with high accuracy.

Secondly, interpolation processing is performed using a plurality of matching reference images to estimate the influence of the deformation of the matching target image, and the deformation amount is obtained from the characteristic value of the matching reference image estimated to be the most matching. By estimating the amount of deformation such as the change of corner angle and the change of radius of the circle,
It has the effect that it can be calculated.

Thirdly, the vertex positions of the corners are determined as 0-dimensional data, that is, points, and the matching degree is defined for 2-dimensional data, that is, in the case of an image, a surface having a certain size or more. It is a statistic. Since the matching degree is calculated in the area including the corner, a large offset due to the corner angle occurs. Therefore, when there is a change in the corner angle of the matching target image, by using the matching method of the present invention, particularly in the calculation of the corner vertex position,
There is an effect that the matching positioning accuracy can be improved.

[Brief description of drawings]

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

FIG. 2 is a diagram showing processing in a second embodiment of the present invention.

FIG. 3 is a diagram showing processing in a third embodiment of the present invention.

[Explanation of symbols]

 1 Matching Reference Image 2 Matching Target Image 3 Correlation Peak Search Mechanism 4 Interpolation System 5 Corner Image with Average Corner Angle b 6 Corner Image with Maximum Corner Angle c 7 Corner Image with Minimum Corner Angle a 8 Corner Angle d Matching target image 9 Image obtained by matching 10 Circular image with average radius e 11 Circular image with maximum radius f 12 Circular image with minimum radius g 13 Matching target image with radius h 14 Image obtained by matching

Claims (12)

[Claims] 1. A plurality of average patterns are provided as a plurality of matching reference images, the matching is performed while shifting the position between each of the plurality of matching reference images and the matching target image, and the result of the matching is performed. As for each of the plurality of matching reference images, the position at which the peak of the matching degree is obtained and the matching degree of the peak are obtained, and a plurality of matching reference images and the matching target image are obtained. A matching degree and a plurality of matching positions including deviations caused by deformation or mismatch between the plurality of matching reference images and the matching target image are obtained, and a plurality of matching positions including the deviations and the plurality of matching degrees are obtained. By correcting the deviation of the plurality of matching positions from
An image matching method for calculating a matching position and a matching degree that do not include a deviation, which is equivalent to that obtained when a pattern having the same shape as the matching target image is used as a matching reference image. 2. A plurality of representative patterns are provided as a plurality of matching reference images, the matching is performed while shifting the position between each of the plurality of matching reference images and the matching target image, and the result of the matching is performed. As for each of the plurality of matching reference images, the position at which the peak of the matching degree is obtained and the matching degree of the peak are obtained, and a plurality of matching reference images and the matching target image are obtained. A matching degree and a plurality of matching positions including deviations caused by deformation or mismatch between the plurality of matching reference images and the matching target image are obtained, and a plurality of matching positions including the deviations and the plurality of matching degrees are obtained. By correcting the deviation of the plurality of matching positions from
An image matching method for calculating a matching position and a matching degree that do not include a deviation, which is equivalent to that obtained when a pattern having the same shape as the matching target image is used as a matching reference image. 3. The matching as a method of obtaining the position where the peak of the matching degree is obtained and the matching degree of the peak uses pixel-based matching and sub-pixel interpolation together, and the plurality of matching degrees are , An estimated peak matching degree obtained by performing the sub-pixel interpolation on the directly obtained matching degree near the peak position of the matching degree directly obtained by the matching, wherein the plurality of matching positions are the estimated peaks. The image matching method according to claim 1 or 2, which is a matching position obtained by performing the sub-pixel interpolation corresponding to the matching degree. 4. The matching is a normal correlation, the plurality of matching positions are correlation positions obtained by performing the normal correlation, and the plurality of matching degrees are correlation values obtained by performing the normal correlation. The image matching method according to any one of claims 1 to 3. 5. The matching is a simple correlation, the plurality of matching positions are correlation positions obtained by performing the simple correlation, and the plurality of matching degrees are correlation substitute values obtained by performing the simple correlation. The image matching method according to any one of claims 1 to 3. 6. A plurality of average patterns are provided as a plurality of matching reference images, the matching is performed while shifting the position between each of the plurality of matching reference images and the matching target image, and the result of the matching is performed. As for each of the plurality of matching reference images, the position at which the peak of the matching degree is obtained and the matching degree of the peak are obtained, and a plurality of matching reference images and the matching target image are obtained. A matching degree and a plurality of matching positions including deviations caused by deformation or mismatch between the plurality of matching reference images and the matching target image are obtained, and a plurality of matching positions including the deviations and the plurality of matching degrees are obtained. By correcting the deviation of the plurality of matching positions from
An image matching method for calculating at least one of a deformation amount, a size, and an angle of the matching target image, which corresponds to that obtained when a pattern having the same shape as the matching target image is used as a matching reference image. 7. A corner image having blur, which is obtained by averaging images near an average corner angle, and a corner image having blur, which is obtained by averaging images near a maximum corner angle, of the plurality of matching reference images. 7. The image matching method according to claim 6, further comprising three images, a corner image having blur that is obtained by averaging images near the minimum corner angle. 8. A plurality of representative patterns are provided as a plurality of matching reference images, the matching is performed while shifting the position between each of the plurality of matching reference images and the matching target image, and the result of the matching is performed. As for each of the plurality of matching reference images, the position at which the peak of the matching degree is obtained and the matching degree of the peak are obtained, and a plurality of matching reference images and the matching target image are obtained. A matching degree and a matching position including a deviation caused by deformation or inconsistency between the plurality of matching reference images and the matching target image are obtained, and from the plurality of matching positions including the deviation and the plurality of matching degrees, Correcting the deviation of the plurality of matching positions,
An image matching method for calculating at least one of a deformation amount, a size, and an angle of the matching target image, which corresponds to that obtained when a pattern having the same shape as the matching target image is used as a matching reference image. 9. The unmatched corner image having a representative corner angle, the unblurred corner image having the maximum corner angle, and the unblurred corner image having the minimum corner angle, the plurality of matching reference images. The image matching method according to claim 8, comprising three images. 10. The matching as a method of obtaining the position where the peak of the matching degree is obtained and the matching degree of the peak uses pixel-based matching and sub-pixel interpolation in combination, and the plurality of matching degrees are equal to each other. , An estimated peak matching degree obtained by performing the sub-pixel interpolation on the directly obtained matching degree near the peak position of the matching degree directly obtained by the matching, wherein the plurality of matching positions are the estimated peaks. The image matching method according to any one of claims 6 to 9, which is a matching position obtained by performing the subpixel interpolation corresponding to the matching degree. 11. A circle image having blur in an edge portion obtained by averaging circle images whose centers coincide with each other near an average size and a circle image having centers coincident with each other near a minimum size. A plurality of matching reference images, three images including a circle image having blur on the edge portion and a circle image having blur on the edge portion obtained by averaging circle images whose centers coincide near the maximum size; The matching is performed while shifting the position between each of the plurality of matching reference images and the matching target image, and as a result of performing the matching, each of the plurality of matching reference images is at the position where the peak of the matching degree is obtained. And a matching degree of the peak, and a plurality of matching degrees between the plurality of matching reference images and the matching target image and a plurality of matching degrees. A plurality of matching positions including deviations caused by deformation or inconsistency between the matching reference image and the matching target image, and the plurality of matching positions based on the plurality of matching positions including the deviations and the plurality of matching degrees. Correct the deviation of
An image matching method for calculating at least one of a deformation amount, a size, or a position of the matching target image, which corresponds to that obtained when a pattern having the same shape as the matching target image is used as a matching reference image. 12. A circle image of a typical size without blurring,
It has three images of a minimum-size unblurred circle image and a maximum-size unblurred circle image as a plurality of matching reference images, and positions each of the plurality of matching reference images and the matching target image. Matching is performed while shifting, and as a result of the matching, a position at which a peak of the matching degree is obtained and a matching degree of the peak are obtained for each of the plurality of matching reference images, and the plurality of matching reference images are obtained. A plurality of matching degrees between the image and the matching target image and a plurality of matching positions including a deviation caused by deformation or mismatch between the plurality of matching reference images and the matching target image are obtained, and the deviation is obtained. Correcting deviations of the plurality of matching positions from the plurality of matching positions including and the plurality of matching degrees,
An image matching method for calculating at least one of a deformation amount, a size, or a position of the matching target image, which corresponds to that obtained when a pattern having the same shape as the matching target image is used as a matching reference image.