JP2010231532A - Image processing program, method and apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent any detection mistake due to the effect of the variation of contrast. <P>SOLUTION: A first calculation means 2 calculates the appearance frequency (degree) of each of the scales of the luminance gradient vectors of an image to be determined (decision object image) 5 whose scale is the same as the scale of an image to be compared (template image) 7 in each of the directions of the luminance gradient vectors. A second calculation means 3 compares each appearance frequency of the decision object image 5 with the appearance frequency of each of the scales of each of the luminance gradient vectors of a template image 7 corresponding to each appearance frequency of the decision object image 5 in each of the directions of the luminance gradient vectors to calculate similarity. A decision means 4 determines whether or not the decision object image 5 is an image which is similar with the template image 7 based on the similarity calculated by the second calculation means 3. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an image processing program, an image processing method, and an image processing apparatus.

In the field of image recognition and measurement, a function for detecting a specific image pattern included in the screen is required.
In order to detect an image pattern, a technique for detecting a similar shape regardless of rotation and scale changes has been developed. For example, in recent years, a technique for estimating a scale (magnification / reduction) parameter of an image feature based on a scale space theory and a technique for estimating a direction parameter of an image pattern from a histogram of a luminance gradient of the image pattern have been developed.

  Thus, it is possible to normalize the scale and rotation of the image pattern to some extent. However, the detection accuracy decreases due to the influence of a minute scale, a difference in rotation, local image pattern deformation, contrast fluctuation, and the like.

  Thus, for example, a method of detecting matching of image patterns by dividing two image patterns to be compared into a plurality of regions, creating histograms in the direction of the luminance gradient vector in each region, and comparing the created histograms It has been known.

David G Lowe, `` Distinctive Image Features from Scale-Invariant Keypoints '', International Journal of Computer Vision, 2004., (America), January 5 2004.

  However, since the magnitude of the brightness gradient vector of the histogram created by the method described above is directly affected by contrast fluctuation, the histogram distribution changes due to local contrast fluctuation, and the same image pattern is obtained. However, the shape of the histogram may be greatly different. For this reason, there is a problem that a detection error occurs.

  Further, in the histogram created by the above-described method, it is not possible to distinguish between the case where there are many small vectors and the case where there are a few large vectors. For this reason, there is still a problem that a detection error occurs.

  The present invention has been made in view of these points, and an object thereof is to provide an image processing program, an image processing method, and an image processing apparatus that prevent detection errors due to the influence of contrast fluctuations.

  In order to achieve the above object, a disclosed image processing program is provided. The image processing program causes the computer to function as a first calculation unit, a second calculation unit, and a determination unit.

The first calculation means calculates an appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the determination target image having the same size as the comparison target image.
The second calculating means generates each appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the comparison target image corresponding to each appearance frequency of the determination target image and each appearance frequency of the determination target image. The similarity is calculated by comparing the respective frequencies.

The determination unit determines whether the determination target image is similar to the comparison target image based on the similarity.
According to this image processing program, the first calculation means determines the appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the determination target image having the same size as the comparison target image. Calculated. By the second calculation means, each appearance frequency of the determination target image and the appearance of each luminance gradient vector for each direction of the luminance gradient vector of the comparison target image corresponding to each appearance frequency of the determination target image The frequency is compared with each other to calculate the similarity. Based on the similarity, the determination unit determines whether or not the determination target image is similar to the comparison target image.

  According to the disclosed image processing program, detection errors due to the influence of contrast fluctuations can be prevented.

It is a figure which shows the outline | summary of the image processing apparatus of embodiment. It is a figure which shows the hardware structural example of an image processing apparatus. It is a block diagram which shows the function of an image processing apparatus. It is a figure explaining the preparation method of a conversion table. It is a figure which shows the flattening of luminance gradient distribution. It is a figure which shows a conversion table. It is a figure which shows the histogram which a histogram preparation part produces. It is a figure explaining the calculation method of similarity. It is a flowchart which shows the process of an image processing apparatus. It is a flowchart which shows the process of an image processing apparatus. It is a figure explaining other calculation methods of similarity.

Hereinafter, embodiments will be described in detail with reference to the drawings.
First, an image processing apparatus according to an embodiment will be described, and then the embodiment will be described more specifically.

FIG. 1 is a diagram illustrating an overview of an image processing apparatus according to an embodiment.
The image processing program according to the embodiment causes the computer (image processing apparatus) 1 to function as the first calculation unit 2, the second calculation unit 3, and the determination unit 4.

  The first calculation means 2 calculates the magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the determination target image (determination target image) 5 having the same size as the comparison target image (template image) 7. The appearance frequency (frequency) for each is calculated.

In FIG. 1, as an example, a histogram 6 indicating the appearance frequency of each luminance gradient vector for each direction (0 ° to 360 °) of the luminance gradient vector of the determination target image 5 is illustrated.
The histogram 6 is one histogram among the determination target images 5 divided into nine regions.

  In the histogram 6, the size is divided into a plurality of sections (size units) for each direction. In FIG. 1, it is divided into four sections. Each section is provided with a pattern 6a that can identify the appearance frequency of the luminance gradient vector. In addition, not only a pattern but a numerical value may be sufficient.

  The second calculation means 3 generates the appearance frequency for each magnitude of the luminance gradient vector for each appearance frequency of the determination target image 5 and the direction of the luminance gradient vector of the template image 7 corresponding to each appearance frequency of the determination target image 5. The similarity is calculated by comparing the respective frequencies.

In FIG. 1, a histogram 8 of the template image 7 is illustrated.
Similarly to the histogram 6, the histogram 8 is divided into a plurality of sections (size units) for each direction. In FIG. 1, it is divided into four sections. Each section is provided with a pattern 8a that can identify the appearance frequency of the luminance gradient vector. In addition, it is not restricted to a pattern, A numerical value may be sufficient.

The histogram 8 is created in advance using a method similar to the method for creating the histogram 6, for example, and stored in a storage means (not shown).
The method for calculating the similarity is not particularly limited. For example, the sum of squares of the frequency differences between the corresponding sections (for example, the bin 6b and the bin 8b in FIG. 1) may be used, which will be described later. The calculation method of “Bhattacharyya” may be used.

The determination unit 4 determines whether the determination target image 5 is an image similar to the template image 7 based on the similarity calculated by the second calculation unit 3.
In FIG. 1, as an example, threshold values are prepared in advance, and the similarity is compared for each of the nine divided areas. When a region having a degree of similarity greater than this threshold is found, it is determined that the determination target image 5 is similar to the template image 7.

  According to the computer 1, by calculating the appearance frequency (frequency) for each magnitude of the luminance gradient vector, a histogram having the same shape can be obtained even for local contrast fluctuations. As a result, detection errors due to the influence of contrast fluctuations can be prevented, and matching performance is improved.

Hereinafter, the embodiment will be described more specifically.
FIG. 2 is a diagram illustrating a hardware configuration example of the image processing apparatus.
The entire image processing apparatus 10 is controlled by a CPU (Central Processing Unit) 101. A random access memory (RAM) 102, a hard disk drive (HDD) 103, a graphic processing device 104, an input interface 105, an external auxiliary storage device 106, and a communication interface 107 are connected to the CPU 101 via a bus 108. Yes.

  The RAM 102 temporarily stores at least part of an OS (Operating System) program and application programs to be executed by the CPU 101. The RAM 102 stores various data necessary for processing by the CPU 101. The HDD 103 stores an OS and application programs. A program file is stored in the HDD 103.

  A monitor 104 a is connected to the graphic processing device 104. The graphic processing device 104 displays an image on the screen of the monitor 104a in accordance with a command from the CPU 101. A keyboard 105 a and a mouse 105 b are connected to the input interface 105. The input interface 105 transmits a signal transmitted from the keyboard 105 a and the mouse 105 b to the CPU 101 via the bus 108.

  The external auxiliary storage device 106 reads information written on the recording medium and writes information on the recording medium. Examples of the recording medium that can be read and written by the external auxiliary storage device 106 include a magnetic recording device, an optical disc, a magneto-optical recording medium, and a semiconductor memory. Examples of the magnetic recording device include an HDD, a flexible disk (FD), and a magnetic tape. Examples of the optical disc include a DVD (Digital Versatile Disc), a DVD-RAM (Random Access Memory), a CD-ROM (Compact Disc Read Only Memory), and a CD-R (Recordable) / RW (ReWritable). Examples of the magneto-optical recording medium include MO (Magneto-Optical disk).

  The communication interface 107 is connected to the network 30. The communication interface 107 transmits and receives data to and from other computers via the network 30.

  With the hardware configuration as described above, the processing functions of the present embodiment can be realized. The following functions are provided in the image processing apparatus 10 having such a hardware configuration.

FIG. 3 is a block diagram illustrating functions of the image processing apparatus.
The image processing apparatus 10 includes a conversion table creation / storage unit 11, a template image histogram creation / storage unit 12, a determination target image histogram creation unit 13, a histogram similarity calculation unit 14, a similarity determination unit 15, and data And an output unit 16.

  The conversion table creation / storage unit 11 includes a conversion table creation unit 111 that creates a conversion table indicating the relationship between the luminance gradient vector and the appearance frequency, and a conversion table storage unit 112 that stores the created conversion table. . Hereinafter, a method for creating a conversion table will be described.

FIG. 4 is a diagram for explaining a conversion table creation method.
First, the conversion table creation unit 111 creates a histogram 22 indicating the appearance frequency for each magnitude of the brightness gradient vector from the plurality of sample images 21.

  The shape of the histogram 22 to be created is not particularly limited because it changes according to the brightness of the sample image 21. However, in the histogram 22 shown in FIG. 4, the appearance frequency increases when the magnitude of the brightness gradient vector is small. Yes.

  Next, the conversion table creation unit 111 flattens the created histogram 22 and divides (reconstructs) the sections so that the appearance frequencies of the sections are equal (sets the unit of the section size). ).

FIG. 5 is a diagram illustrating flattening of the luminance gradient distribution.
In FIG. 5, as an example, it is assumed that the magnitude of the brightness gradient vector of the created histogram 22 is 10 sections.

As shown in FIG. 5, the appearance frequency of each section is different. A flattened graph 23 is generated by dividing the section so that the appearance frequencies of the sections become equal.
For example, the appearance frequency of section 1 is taken out as it is and is set as section 1 of the flattened graph 23. Next, the appearance frequency of the section 2 and the appearance frequency of the section 3 are added to obtain the section 2 of the flattened graph 23. Next, the appearance frequencies of the sections 4 to 6 are added to form a section 3 of the flattened graph 23. Next, the appearance frequencies of the sections 7 to 10 are added to obtain a section 4 of the flattened graph 23. As a result, the appearance frequency of each section is “128”. By this processing, the information amount in each section becomes (almost) equal, and a comparison result with higher accuracy can be obtained in the comparison of similarity described later.

  In FIG. 5, the appearance frequencies of the respective sections of the flattened graph 23 are all equal, but they may not be all equal, and the sections are divided so as to be the most flat (the existence probabilities are equal). You may make it do.

Thereafter, the conversion table creation unit 111 creates a conversion table for converting the magnitude of the luminance gradient vector into the section number of the flattened graph 23.
FIG. 6 is a diagram showing a conversion table.

The conversion table 11a has columns of size and section.
In the size column, a number indicating the size of the luminance gradient vector is set. This number corresponds to the number on the horizontal axis of the histogram 22 shown in FIG.

In the section column, a number indicating the section of the flattening graph 23 is set.
Returning to FIG. 4, the description will be continued.
The template image histogram creation / storage unit 12 includes a template image storage unit 121, a brightness gradient vector calculation unit 122, a histogram creation unit 123, and a histogram storage unit 124.

The template image storage unit 121 stores an image (template image) of a pattern to be detected.
The luminance gradient vector calculation unit 122 performs a differentiation process on the template image stored in the template image storage unit 121 to calculate the luminance gradient vector of each pixel.

  The histogram creation unit 123 divides the template image into a plurality of regions. Then, the luminance gradient vector is divided by direction and size for each region. Furthermore, the size is assigned to each section. This creates a two-dimensional histogram in which the appearance frequency for each magnitude of the brightness gradient vector is described for each direction of the brightness gradient vector.

  Here, when creating the histogram, the conversion table 11a stored in the conversion table storage unit 112 (created by the conversion table creation unit 111) is referred to. Then, the appearance frequency is accumulated for each section of the conversion table 11a.

The histogram creation unit 123 stores the created histogram in the histogram storage unit 124.
FIG. 7 is a diagram illustrating a histogram created by the histogram creation unit.

The image pattern 50 of the template image is divided into nine regions R ₀₀ to R ₂₂ . The histogram of the brightness gradient is created for each of the regions R ₀₀ to R ₂₂ . That is, nine brightness gradient histograms are created. 7 illustrates a histogram HR ₀₀ regions R _00.

Histogram HR _00, the vertical axis shows the magnitude of the brightness gradient and the horizontal axis represents each direction.
Further, the numerical value in each bin of the histogram HR ₀₀ indicates the appearance frequency of the section in that direction. For example, the bin b1 in the section 2 in the direction 0 ° indicates that the appearance frequency is “6”.

Returning to FIG. 4, the description will be continued.
The determination target image histogram creation unit 13 includes an image reading unit 131, a read image storage unit 132, a region image extraction unit 133, a luminance gradient vector calculation unit 134, and a histogram creation unit 135.

The image reading unit 131 reads, for example, an image captured by a camera or an image selected by a user as an image whose similarity is to be determined, and converts it into digital data.
The read image storage unit 132 stores the digital data converted by the image reading unit 131.

  The region image extraction unit 133 extracts an image region having the same size as the template image as a candidate region from the designated location in the image data of the entire screen stored in the read image storage unit 132.

The luminance gradient vector calculation unit 134 performs a differentiation process on the candidate region image extracted by the region image extraction unit 133 to calculate the luminance gradient vector of each pixel.
The histogram creation unit 135 divides the candidate region image extracted by the region image extraction unit 133 to create a two-dimensional histogram of the direction and size of the gradient vector. Similarly to the histogram creation unit 123, the conversion table 11a stored in the conversion table storage unit 112 is referred to when creating this histogram. Then, the appearance frequency is accumulated for each section of the conversion table 11a. Thereby, the section width of the histogram created by the histogram creation unit 123 is equal to the section width of the histogram created by the histogram creation unit 135.

  The histogram similarity calculation unit 14 calculates the similarity between the histogram of the template image (calculated by the histogram creation unit 123) stored in the histogram storage unit 124 and the histogram of the candidate area image calculated by the histogram creation unit 135. . Specifically, the similarity is calculated by comparing the appearance frequencies of the template image and the candidate area in the same direction and in the same section.

FIG. 8 is a diagram for explaining a method of calculating similarity.
In the calculation of the similarity, each region R ₀₀ to region R ₂₂ of the histogram is expressed by a one-dimensional vector as shown in FIG. Specifically, first, the bins in the section 1 of the region R ₀₀ are arranged in order of direction. Then, the bins in the section 2 are arranged in the order of direction after the bins in the direction 315 ° of the section 1. When the bins of the section 4 of the region R ₀₀ are arranged, the bins of the section 1 of the region R ₀₁ are arranged. Thereafter, the bins are arranged in the same manner, and finally, the bins of the section 4 of the region R ₂₂ are arranged in the order of direction.

Similar processing is performed for the histogram of the corresponding template image.
The similarity is calculated using the similarity calculation method of “Bhattacharyya” according to the following equation (1).

Here, p (i) and q (i) are the appearance frequencies of a one-dimensional histogram for the template image and the candidate region image, respectively, and N is the number of elements of the one-dimensional vector.
The value of similarity B in equation (1) is a real value in the range of “0.0 to 1.0”, and a larger value is obtained as the similarity is higher.

The similarity determination unit 15 determines pattern matching from the similarity B calculated by the histogram similarity calculation unit 14 by, for example, threshold processing. Specifically, by preparing a threshold value in advance, and the vector was one-dimensional histograms of the regions R ₀₀ ~ region R ₂₂ of the template image, one-dimensional histograms of the regions R ₀₀ ~ region R ₂₂ of the candidate regions If the similarity B calculated with the converted vector is greater than the threshold, it is determined that the image of the candidate area is similar to the template image.

The data output unit 16 outputs the determination result determined by the similarity determination unit 15 to the monitor 104a.
Next, processing of the image processing apparatus 10 will be described.

9 and 10 are flowcharts showing processing of the image processing apparatus.
First, the conversion table 11a from the magnitude of the brightness gradient vector of the sample image to the section of the histogram is created (step S1).

Next, the template image storage unit 121 stores a template image (step S2).
Next, the luminance gradient vector calculation unit 122 calculates the luminance gradient vector of the template image (step S3).

Next, the histogram creation unit 123 calculates a histogram of the direction and size of the brightness gradient vector of the template image (step S4).
Next, the image reading unit 131 reads an image to be determined (step S5).

Then, the read image is read and stored in the image storage unit 132 (step S6).
Next, the region image extraction unit 133 extracts an image that is a candidate region of the image stored in the read image storage unit 132 (step S7 in FIG. 10).

Next, the luminance gradient vector calculation unit 134 calculates the luminance gradient vector of the candidate region image (step S8).
Next, the histogram creation unit 135 calculates a histogram of the direction and size of the luminance gradient vector of the image of the candidate area (step S9).

  Next, the histogram similarity calculation unit 14 calculates the similarity between the brightness gradient vector histogram of the template image and the brightness gradient vector histogram of the candidate region image (step S10). Then, the similarity determination unit 15 determines whether or not the similarity is greater than a predetermined threshold (step S11).

  When the similarity is larger than a predetermined threshold (Yes in step S11), it is determined that the two images are the same. Then, the data output unit 16 outputs the determination result to the monitor 104a.

  On the other hand, when the similarity is equal to or less than a predetermined threshold (No in step S11), the similarity determination unit 15 determines whether or not the processing in steps S7 to S11 has been performed for all candidate regions ( Step S12).

  When the processing of step S7 to step S11 is not performed for all candidate regions (No in step S12), the process proceeds to step S7, and the designated image data in the full screen image data stored in the read image storage unit 132 is designated. Another image area having the same size as the template image is extracted from the place as a candidate area (step S7). And the process after step S8 is performed repeatedly.

  On the other hand, when the process of step S7 to step S11 is performed for all candidate regions (Yes in step S12), it is determined that the image to be determined is not similar to the template image. Then, the data output unit 16 outputs the determination result to the monitor 104a.

Above, description of the process of the image processing apparatus 10 is complete | finished.
As described above, according to the image processing apparatus 10, the histogram creation unit 123 and the histogram creation unit 135 each create a two-dimensional histogram in which the appearance frequency of the luminance gradient vector is described for each section. Therefore, it is possible to obtain a histogram having the same shape with respect to local contrast fluctuations.

Thereby, the histogram similarity calculation part 14 can calculate the similarity B more correctly. For this reason, matching performance improves.
Further, the magnitude of the brightness gradient vector is divided by the section width so that the existence probability of the magnitude of the brightness gradient vector in each section is equal. Thereby, since the average information amount (average entropy) which the histogram about the magnitude | size of a gradient vector has can be maximized, matching performance can be improved further.

  In the present embodiment, the similarity determination unit 15 determines image matching by threshold processing from the similarity B calculated by the histogram similarity calculation unit 14. However, the present invention is not limited to this. For example, the degree of similarity is calculated with a low-dimensional vector having the sum of the frequency in the angular direction as an element for the size-direction section of each region. Then, only when the degree of similarity is greater than or equal to a threshold value, the degree of similarity of a vector composed of all the elements of the histogram may be calculated / determined. Thereby, the calculation amount of similarity calculation / determination can be reduced.

FIG. 11 is a diagram for explaining another method of calculating similarity.
As shown in FIG. 11, first, the value "44" obtained by adding all the frequency intervals 1 region R _00, numerical values obtained by adding all the frequency interval 2 regions R ₀₀ "159", ... , value "60" obtained by adding all the frequency of appearance of the section 3 of the area R _22, obtains the value "129" obtained by adding all the frequency interval fourth region R _22. This is a low-dimensional vector. Similar processing is performed for the histogram of the corresponding template image. Then, using these vectors, for example, the similarity is calculated by the similarity calculation method of “Bhattacharyya”.

  Note that the processing performed by the image processing device 10 may be distributed by a plurality of devices. For example, another device may calculate the histogram of the template image to create a histogram, and the image processing device may use the histogram to determine the degree of similarity with the determination target image. Good.

  The image processing program, the image processing method, and the image processing apparatus according to the present invention have been described based on the illustrated embodiment. However, the present invention is not limited to this, and the configuration of each unit has the same function. Can be replaced with any structure having Moreover, other arbitrary structures and processes may be added to the present invention.

In addition, the present invention may be a combination of any two or more configurations (features) of the above-described embodiments.
The above processing functions can be realized by a computer. In that case, a program describing the processing contents of the functions of the image processing apparatus 10 is provided. By executing the program on a computer, the above processing functions are realized on the computer. The program describing the processing contents can be recorded on a computer-readable recording medium. Examples of the computer-readable recording medium include a magnetic recording device, an optical disk, a magneto-optical recording medium, and a semiconductor memory. Examples of the magnetic recording device include a hard disk device (HDD), a flexible disk (FD), and a magnetic tape. Examples of the optical disc include a DVD (Digital Versatile Disc), a DVD-RAM (Random Access Memory), a CD-ROM (Compact Disc Read Only Memory), and a CD-R (Recordable) / RW (ReWritable). Examples of the magneto-optical recording medium include MO (Magneto-Optical disk).

  When distributing the program, for example, a portable recording medium such as a DVD or a CD-ROM in which the program is recorded is sold. It is also possible to store the program in a storage device of a server computer and transfer the program from the server computer to another computer via a network.

  A computer that executes an image processing program stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. Then, the computer reads the program from its own storage device and executes processing according to the program. The computer can also read the program directly from the portable recording medium and execute processing according to the program. In addition, each time the program is transferred from the server computer, the computer can sequentially execute processing according to the received program.

1 Computer (image processing device)
DESCRIPTION OF SYMBOLS 2 1st calculation means 3 2nd calculation means 4 Determination means 5 Determination object image 6, 8 Histogram 6a, 8a Pattern 6b, 8b, b1 Bin 7 Template image 10 Image processing apparatus 11 Conversion table preparation and storage part 11a Conversion table DESCRIPTION OF SYMBOLS 12 Template image histogram preparation / memory | storage part 13 Determination object image histogram preparation part 14 Histogram similarity calculation part 15 Similarity determination part 16 Data output part 21 Sample image 22 Histogram 23 Flattening graph 50 Image pattern 111 Conversion table preparation part 112 Conversion table Storage unit 121 Template image storage unit 122, 134 Brightness gradient vector calculation unit 123, 135 Histogram creation unit 124 Histogram storage unit 131 Image reading unit 132 Read image storage unit 133 Area image extraction unit H R ₀₀ histogram R _{00 to} R ₂₂ region

Claims (7)

Computer
First calculating means for calculating an appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the determination target image having the same size as the size of the image to be compared;
The appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the comparison target image corresponding to the appearance frequency of the determination target image and the appearance frequency of the determination target image And a second calculating means for calculating the similarity by comparing each of
Determination means for determining whether the image to be determined is an image similar to the image to be compared based on the similarity;
An image processing program that functions as an image processing program. Causing the computer to function as a dividing unit that divides the image to be determined into a plurality of regions;
The first calculation means calculates the appearance frequency for each area, and the second calculation means divides the comparison target image into a plurality of areas, and the appearance calculated for each of the areas is calculated. The image processing program according to claim 1, wherein the frequency is compared with the appearance frequency calculated by the first calculation unit. The first calculation means summarizes the appearance frequencies of the luminance gradient vectors in different magnitude units according to the magnitude of the luminance gradient vector,
The second calculation means calculates the similarity by comparing the appearance frequencies of the comparison target image and the determination target image in the same direction and in the same unit, respectively. Item 8. The image processing program according to Item 1.   4. The image processing program according to claim 3, wherein the first calculation means sets the unit so that the number of the luminance gradient vectors becomes equal.   The image processing program according to claim 1, wherein the determination unit determines that the image pattern having the similarity greater than a threshold prepared in advance is an image similar to the template image. Computer
For each direction of the luminance gradient vector of the determination target image having the same size as the size of the image to be compared, calculate the appearance frequency for each size of the luminance gradient vector,
The appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the comparison target image corresponding to the appearance frequency of the determination target image and the appearance frequency of the determination target image To calculate the similarity,
Determining whether the image to be determined is an image similar to the image to be compared based on the similarity;
An image processing method. A first calculation unit that calculates an appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the determination target image having the same size as the comparison target image;
The appearance frequency for each magnitude of the luminance gradient vector for each direction of the luminance gradient vector of the comparison target image corresponding to the appearance frequency of the determination target image and the appearance frequency of the determination target image And a second calculation unit that calculates the similarity by comparing each of
A determination unit configured to determine whether the image to be determined is an image similar to the image to be compared based on the similarity;
An image processing apparatus comprising:

Images