Image processing device, image processing method, and image processing program

Abstract

The present invention has the object of improving accuracy of image processing which determines whether or not an image relating to a predetermined object is included in a target image. An image processing device 1 comprises: a sample distribution index calculation unit 25 that calculates a sample distribution index indicating variations of a plurality of sample features extracted from the sample image S1 relating to a predetermined object in the sample image S; a matching distribution index calculation unit 26 that calculates a matching distribution index indicating variations, in the sample image S1, of a plurality of matching features which are out of the plurality of sample features and match a plurality of query features extracted from a query image Q1 or Q2, which is a target of determining whether or not the sample image S1 is included; and an exclusion determination unit 27 that determines that the sample image S1 is not included in the query image Q1 or Q2 when a difference between the sample distribution index and the matching distribution index exceeds a predetermined criterion.

Description

Image processing device, image processing method, and image processing program

[0001] The present disclosure relates to image processing techniques.

[0002] Conventionally, a similar region detection device has been proposed that includes an acquisition unit that acquires a first image and a second image, a feature point extraction unit that extracts feature points from each of the first image and the second image, a matching unit that matches the feature points extracted from the first image with the feature points extracted from the second image to detect corresponding points between the images, an outermost contour extraction unit that extracts an outermost contour from each of the first image and the second image, and a detection unit that detects similar regions, which are partial regions that are similar between the first image and the second image, from each of the first image and the second image based on the outermost contour and the number of corresponding points (see Patent Document 1).

[0003] In addition, various techniques have been proposed for matching images by collating features extracted from images (see Non-Patent Documents 1 and 2).

[0004] Japanese Patent Application Laid-Open No. 2021-051581

[0005] “OpenCV Feature Matching”, [online], OpenCV, [searched on February 27, 2020], Internet <URL: https: / / docs.opencv.org / 4.x / dc / dc3 / tutorial_py_matcher.html>Paul-Edouard Sarlin, 3 others, “SuperGlue: Learning Feature Matching with Graph Neural Networks”, CVPR2020

[0006] In the past, various techniques have been proposed for recognizing predetermined targets (imaged objects, characters, etc.) contained in images, but there is room for improvement in the accuracy of recognizing targets in images. In view of the above-mentioned problems, an object of the present disclosure is to improve the accuracy of recognizing predetermined targets contained in images.

[0007] An example of the present disclosure is an image processing device including: a sample distribution index calculation means for calculating a sample distribution index indicating the variation in a sample image of a plurality of sample features extracted from the sample image relating to a specified object; a match distribution index calculation means for calculating a match distribution index indicating the variation in the sample image of a plurality of match features that match a plurality of query features extracted from a query image that is to be determined as to whether or not the query image includes an image relating to the specified object, among the plurality of sample features; and an exclusion determination means for determining that the query image does not include an image relating to the specified object if a difference between the sample distribution index and the match distribution index exceeds a specified criterion.

[0008] The present disclosure can be understood as an image processing device, a system, a method executed by a computer, or a program executed by a computer. The present disclosure can also be understood as such a program recorded on a recording medium readable by a computer or other device, machine, etc. Here, a recording medium readable by a computer, etc. refers to a recording medium that stores information such as data and programs by electrical, magnetic, optical, mechanical, or chemical action and can be read by a computer, etc.

[0009] According to the present disclosure, it is possible to improve the accuracy of recognizing a predetermined object contained in an image.

[0010] 1 is a schematic diagram showing the configuration of a system according to an embodiment; FIG. 2 is a diagram showing an outline of the functional configuration of an image processing device according to an embodiment; FIG. 3 is a diagram showing an example of a query image according to an embodiment; FIG. 4 is a diagram showing an example of a sample image related to an object and features extracted from the sample image according to an embodiment; FIG. 5 is a diagram showing an outline of calculation of the density of sample features according to an embodiment; FIG. 6 is a diagram showing an outline of calculation of the density of match features according to an embodiment; FIG. 7 is a diagram showing an outline of comparison of the standard deviation of feature densities for each of a sample image and a query image according to an embodiment; FIG. 8 is a diagram showing the relationship between the determination result by an object determination unit, the standard deviation of feature densities, and the difference in standard deviations for each of a sample image and a query image according to an embodiment; and FIG. 9 is a flowchart showing the flow of inspection processing according to an embodiment. FIG. 10 is a diagram showing an outline of the functional configuration of an image processing device related to a variation; and FIG. 11 is a flowchart showing the flow of inspection processing related to a variation.

[0011] Hereinafter, embodiments of an image processing device, method, and program according to the present disclosure will be described with reference to the drawings. However, the embodiments described below are merely examples, and the image processing device, method, and program according to the present disclosure are not limited to the specific configurations described below. In implementing the present disclosure, specific configurations according to the embodiment may be appropriately adopted, and various improvements and modifications may be made.

[0012] Conventionally, feature matching has been performed to compare images, and various image processing techniques have been proposed to determine whether a target image contains an image related to a specific object. However, while these conventional techniques are effective in detecting similar regions, which are partial regions that are similar to each other between images, there is room for improvement in terms of performing matching that takes into consideration the positional relationships and distribution conditions between features in a huge number of feature groups.

[0013] In view of the above-mentioned problems, an object of the present disclosure is to improve the accuracy of image processing for determining whether or not a target image contains an image relating to a predetermined object.

[0014] The image processing device, method, and program according to this embodiment make it possible to perform matching that takes into greater consideration the positional relationships and distribution conditions between features in a feature group, thereby improving the accuracy of image processing for determining whether a target image contains an image related to a specified object.

[0015] In this embodiment, an embodiment in which the technology according to the present disclosure is implemented to confirm whether an advertisement as a predetermined object has been installed correctly in accordance with instructions for displaying the advertisement will be described. However, the technology according to the present disclosure can be widely used to determine whether an image of a target contains an image related to a predetermined object, and the application of the present disclosure is not limited to the examples shown in the embodiment. For example, the object may not be a tangible object, but may be an object drawn in a virtual space or an object drawn using drawing software. Furthermore, for example, the image may not be an optically captured image, but may be an image drawn in a virtual space or an image drawn using drawing software.

[0016] 1 is a schematic diagram showing the configuration of a system according to this embodiment. The system according to this embodiment includes an image processing device 1, an imaging device 81, and a user terminal 9, which are connected to a network and can communicate with each other.

[0017] The image processing device 1 is a computer including a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage device 14 such as an EEPROM (Electrically Erasable and Programmable Read Only Memory) or an HDD (Hard Disk Drive), and a communication unit 15 such as a NIC (Network Interface Card). However, the specific hardware configuration of the image processing device 1 can be omitted, replaced, or added as appropriate depending on the embodiment. Furthermore, the image processing device 1 is not limited to a device consisting of a single housing. The image processing device 1 may be realized by a plurality of devices using so-called cloud or distributed computing technology.

[0018] The imaging device 81 captures an image of the inspection target to obtain a query image, which will be described later. The imaging device 81 may be a general digital camera or other device capable of recording light incident from the target, and its specific configuration is not limited. In this embodiment, the inspection target is a store in which an advertisement is installed as a predetermined object.

[0019] The user terminal 9 is a terminal device used by a user. The user terminal 9 is a computer equipped with a CPU, ROM, RAM, a storage device, a communication unit, an input device, an output device, etc. (not shown). However, the specific hardware configuration of the user terminal 9 can be omitted, replaced, or added as appropriate depending on the embodiment. Furthermore, the user terminal 9 is not limited to a device consisting of a single housing. The user terminal 9 may be realized by multiple devices using so-called cloud or distributed computing technology, etc. Users use various services provided by the system according to this embodiment via these user terminals 9.

[0020] 2 is a diagram illustrating an outline of the functional configuration of the image processing device 1 according to this embodiment. The image processing device 1 functions as an image processing device including a query image acquisition unit 21, a feature extraction unit 22, a match feature identification unit 23, an object determination unit 24, a sample distribution index calculation unit 25, a match distribution index calculation unit 26, an exclusion determination unit 27, and a notification unit 28, by a program recorded in the storage device 14 being read into the RAM 13 and executed by the CPU 11, which controls each piece of hardware included in the image processing device 1. Note that in this embodiment and other embodiments described below, each function included in the image processing device 1 is executed by the CPU 11, which is a general-purpose processor, but some or all of these functions may be executed by one or more dedicated processors.

[0021] The query image acquisition unit 21 acquires a query image to be inspected, in which an installed object (in this embodiment, an advertisement) and a target where the object is installed (in this embodiment, a store) are simultaneously captured. Although the method for acquiring the query image is not limited, in this embodiment, an example will be described in which a query image captured using the imaging device 81 is acquired via the user terminal 9.

[0022] 3 is a diagram illustrating an example of a query image in this embodiment, which shows a query image Q1 in which an advertisement A1 is captured together with a store, and a query image Q2 in which an advertisement A2, which is an advertisement for the same product or service as advertisement A1 but has a different layout, is captured together with the store.

[0023] As described above, in this embodiment, an embodiment will be described in which the technology according to the present disclosure is implemented to check whether an advertisement as a predetermined object has been placed correctly in accordance with instructions for displaying the advertisement. When a company displays advertisements in a store or the like, even if the products or services being advertised are the same, detailed display rules (such as display location and display time) may be defined for multiple types of advertisements prepared for the advertisements in order to achieve the intended effect. For example, instructions are given for the size, layout, and content of the advertisements as to where they should be displayed, such as on a signboard outside the store, at the store entrance, on a product shelf or sideboard inside the store.

[0024] A person in charge of checking whether the advertisement has been placed in accordance with the instructions for displaying the advertisement takes a photograph at the location where the advertisement is displayed (e.g., a store) so that the advertisement and its installation location are included in the same image, and this photograph is used as a query image.

[0025] The feature extraction unit 22 extracts a plurality of features (feature points; hereinafter, the features extracted from the query image are referred to as "query features") from a query image Q1 or Q2, which is the target of determination as to whether or not an image relating to a predetermined object is included. In this embodiment, the feature extraction unit 22 further extracts a plurality of features (hereinafter, the features extracted from the sample image are referred to as "sample features") from a sample image S1 relating to the predetermined object. Note that the types of algorithms and descriptors used in extracting features from an image are not limited, and currently known algorithms and descriptors as well as any algorithms and descriptors developed in the future may be used.

[0026] 4 is a diagram showing a sample image S1 of an object according to this embodiment and examples of features extracted from the sample image S1. This diagram shows the sample image S1 of the advertisement A1 described above. Since the object according to this embodiment is an advertisement to be displayed in a store, the sample image may be advertisement production data (e.g., printing data), or photographic data obtained by capturing an image of an actually produced advertisement.

[0027] The "+" in the figure indicates the position of the sample feature extracted from the sample image S1. Feature extraction from the query image Q1 or Q2 is not shown in the figure because it is generally similar to feature extraction from the sample image, except that the image input to the extractor is a query image rather than a sample image.

[0028] The match feature identification unit 23 determines whether each of the multiple sample features extracted from the sample image S1 matches the multiple query features extracted from the query image Q1 or Q2, and thereby identifies multiple sample features (hereinafter referred to as ``match features'') that match any of the multiple query features among the multiple sample features.

[0029] The object determination unit 24 determines whether or not the query image Q1 or Q2 contains an image related to a predetermined object, based on the plurality of match features, using a method different from that of the exclusion determination unit 27, which will be described later. Here, the specific method by which the object determination unit 24 determines whether or not the query image Q1 or Q2 contains an image related to a predetermined object may be a method different from that of the exclusion determination unit 27, which will be described later, and the specific determination method is not limited. Note that in the present embodiment, the match feature identification unit 23 and the object determination unit 24 identify the plurality of match features using a trained machine learning model and determine whether or not the query image Q1 or Q2 contains an image related to a predetermined object.

[0030] Specifically, for example, the match feature identification unit 23 and the object determination unit 24 can determine whether the query image Q1 or Q2 contains the sample image S1 by using a method of matching features by attempting to optimize the transformation of the features (homography matrix in the case of a planar object, or fundamental matrix in the case of a three-dimensional object) between the sample image S1 and the query image Q1 or Q2, and determining that matching is established if optimization is possible (if the transformation or matrix is ​​valid). Note that a more specific matching method is a conventional technique as shown in Non-Patent Documents 1 and 2, and therefore a description thereof will be omitted.

[0031] The sample distribution index calculation unit 25 calculates a sample distribution index that indicates the variation in the sample image S1 of a plurality of sample features extracted from the sample image S1 related to a predetermined object. In this embodiment, the sample distribution index calculation unit 25 calculates the density of the sample features included in each grid when the sample image S1 is divided into a plurality of grids, and calculates the standard deviation of the density of the sample features included in each grid, thereby calculating the sample distribution index. Note that in this embodiment, an example will be described in which the standard deviation is used as the distribution index, but a known statistical quantity that indicates the variation from the mean value, such as variance, may also be used as the distribution index.

[0032] 5 is a diagram showing an outline of calculation of the density of sample features in this embodiment. In this embodiment, the sample image S1 is divided into 16 grids arranged in a matrix of 4 rows and 4 columns, and the density of sample features for each grid is calculated. Specifically, the sample distribution index calculation unit 25 counts the number of features included in each grid and calculates the ratio to the number of features included in the entire image, thereby calculating the density of features for each grid. That is, in this embodiment, the feature density d of grid i is ｉ is calculated according to the following formula:

[0033] Then, the sample distribution index calculation unit 25 calculates the standard deviation of the feature density of the sample image S1 according to the calculated feature density for each grid. That is, in this embodiment, the standard deviation sd of the feature density of the sample image S1 is calculated according to the following formula: where n is the number of grids (in this embodiment, n=16), d μ is the average density.

[0034] The match distribution index calculation unit 26 calculates a match distribution index indicating the variation in the sample image S1 of a plurality of match features that match a plurality of query features extracted from the query image Q1 or Q2, among a plurality of sample features. In this embodiment, the match distribution index calculation unit 26 calculates the density of match features included in each grid when the sample image S1 is divided into a plurality of grids, and calculates the standard deviation of the density of match features included in each grid to calculate the match distribution index.

[0035] 6 is a diagram showing an overview of calculation of the density of match features in this embodiment. The example shown in the upper row shows the density of match features for each grid that matched with the query image Q1 illustrated in FIG. 3, and the example shown in the lower row shows the density of match features for each grid that matched with the query image Q2. For the match distribution index, the sample image S1 is equally divided into 16 grids arranged in a 4-row, 4-column matrix, and the density of match features for each grid is calculated. A more specific method for calculating the match distribution index (standard deviation of the density of match features) is similar to the method for calculating the sample distribution index described above, and therefore will not be described here.

[0036] 7 is a diagram illustrating an overview of comparing the standard deviations of feature densities for sample image S1 and query images Q1 and Q2 in this embodiment. Because this diagram is intended to explain the overview of comparing standard deviations, numerical values ​​are omitted for the scales of the vertical and horizontal axes. In this diagram, the vertical axis represents the feature density of the grid, and the horizontal axis represents the deviation from the average value of the feature density, with the center of the horizontal axis representing the average value. This diagram shows that the difference between the standard deviations of sample image S1 and query image Q2 is relatively large compared to the difference between the standard deviations of sample image S1 and query image Q1.

[0037] The exclusion determination unit 27 determines that the query image Q1 or Q2 does not include an image related to a predetermined object (sample image S1) when the difference between the sample distribution index and the match distribution index exceeds a predetermined standard. The exclusion determination unit 27 determines that the query image Q1 or Q2 does not include an image related to a predetermined object when the absolute value of the difference between the standard deviation as the sample distribution index and the standard deviation as the match distribution index exceeds a predetermined threshold, regardless of the determination result by the object determination unit 24. That is, in this embodiment, the exclusion determination unit 27 is a functional unit for determining whether the determination result by the object determination unit 24 is correct or incorrect.

[0038] 8 is a diagram showing the relationship between the determination results by the object determination unit 24, the standard deviations of feature densities, and the difference in standard deviations for each of the sample image S1, query images Q1, and Q2 in this embodiment. As can be seen from FIGS. 3 and 4 , advertisement A1 associated with sample image S1 is captured in query image Q1, while advertisement A1 associated with sample image S1 is not captured in query image Q2. However, advertisement A2 captured in query image Q2 differs from advertisement A1 associated with sample image S1 only in layout, with common elements. Therefore, the object determination unit 24 according to this embodiment determines that advertisement A1 associated with sample image S1 is captured in both query image Q1 and query image Q2 (i.e., the determination result of query image Q2 by the object determination unit 24 is incorrect).

[0039] Therefore, the exclusion determination unit 27 compares the distribution index (standard deviation) of the sample features calculated by the sample distribution index calculation unit 25 with the distribution index (standard deviation) of the match features for each of the query image Q1 and the query image Q2 calculated by the match distribution index calculation unit 26. As a result, it is found that the difference between the sample distribution index and the match distribution index for the query image Q2 is relatively large compared to the difference for the query image Q1. Because the difference for the query image Q2 exceeds a predetermined threshold, the exclusion determination unit 27 determines that the advertisement for the sample image S1 has not been captured for the query image Q2, regardless of the determination result by the object determination unit 24.

[0040] The notification unit 28 notifies the user of the result of the determination by the object determination unit 24 and / or the result of the determination by the exclusion determination unit 27 .

[0041] <Processing Flow> Next, the processing flow executed by the image processing device according to this embodiment will be described. Note that the specific content and processing order of the processing described below are an example for implementing the present disclosure. The specific processing content and processing order may be selected as appropriate depending on the embodiment of the present disclosure.

[0042] 9 is a flowchart showing the flow of the inspection process according to this embodiment. The process shown in this flowchart is executed when an instruction to start an inspection from a user is accepted.

[0043] In the inspection process described in this flowchart, a first matching process (step S1) is first executed, in which a first matching of a feature group of a query image and a predetermined sample image is performed based on a trained model. Then, the result of the first matching process is determined (step S2). If the result of the first matching indicates that the query image is similar to the sample image (YES in step S2), a second matching process (step S3) is executed, in which a second matching of the distribution of the feature group of the query image is performed. Once the results of the first matching process and / or the second matching process are obtained, the user is notified of the installation state determination result (step S4). The notification unit 28 notifies the user of the determination result in step S1 and / or the determination result in step S3. The process shown in this flowchart then ends.

[0044] The processes (steps S101 to S105) included in the first matching process (step S1) will be described in detail below.

[0045] In step S101, a query image is acquired. The worker uses the imaging device 81 to capture an image of the inspection target (in this embodiment, a store in which an advertisement is displayed as a predetermined object) and inputs image data of the obtained query image into the image processing device 1. The imaging method and the method of inputting the image data into the image processing device 1 are not limited, but in this embodiment, the inspection target is captured using the imaging device 81, and the image data transferred from the imaging device 81 to the user terminal 9 via communication or a recording medium is further transferred to the image processing device 1 via a network, thereby inputting the image data of the query image into the image processing device 1. When the query image is acquired by the query image acquisition unit 21, the process proceeds to step S102.

[0046] In steps S102 and S103, features related to the query image and features related to the sample image are acquired. The feature extraction unit 22 extracts query features from the query image acquired in step S101 (step S102). Then, the match feature identification unit 23 acquires sample features extracted from the sample image (step S103). Note that the sample features acquired in step S103 may be those extracted and stored in advance by the feature extraction unit 22 or the like before the start of the processing related to this flowchart, or may be those extracted by the feature extraction unit 22 during the processing related to this flowchart. Then, the processing proceeds to step S104.

[0047] In steps S104 and S105, it is determined whether the query image contains a sample image. The match feature identification unit 23 determines whether each of the sample features acquired in step S103 matches the query feature extracted in step S102, thereby identifying a match feature among the sample features that matches any of the query features (step S104). Then, the object determination unit 24 determines whether the query image contains an image related to a predetermined object based on the matching result in step S104 (step S105). As described above, in this embodiment, a method for determining whether the sample image and the query image match by using a method for determining whether feature conversion between the sample image and the query image can be optimized is adopted. However, the matching method adopted in the first matching process may be different from the second matching process described below, and the matching method adopted in the first matching process is not limited to the example shown in this disclosure. The first matching process then ends.

[0048] The processes (steps S301 to S304) included in the second matching process (step S3) will be described in detail below.

[0049] In steps S301 and S302, the standard deviations of the sample features and the match features included in the multiple grids into which the sample image is divided are acquired. The exclusion determination unit 27 acquires the standard deviation of the density of the sample features included in each grid when the sample image is divided into multiple grids (16 grids arranged in a 4-row, 4-column matrix in this embodiment) (step S301). The standard deviation of the sample features acquired in step S301 may be calculated and stored by the sample distribution index calculation unit 25 or the like before the start of the processing related to this flowchart, or may be calculated by the sample distribution index calculation unit 25 during the processing related to this flowchart. The match distribution index calculation unit 26 calculates the density of the match features included in each grid when the sample image is divided into multiple grids (16 grids arranged in a 4-row, 4-column matrix in this embodiment), and calculates the standard deviation of the density of the match features included in each grid (step S302). The processing then proceeds to step S303.

[0050] In steps S303 and S304, it is verified whether the query image contains a sample image based on the standard deviation. The exclusion determination unit 27 calculates the difference between the standard deviation of the sample feature and the standard deviation of the match feature, and compares the calculated difference with a preset threshold (step S303). If the result of the comparison shows that the calculated difference exceeds the threshold, the exclusion determination unit 27 determines that the query image acquired in step S101 does not contain an object related to the sample image (step S304). Then, the second matching process ends.

[0051] That is, according to the process shown in this flowchart, if the result of the first matching process is positive (indicating that the query image is similar to the sample image), a second matching process is executed, and the judgment result of the first matching process is verified based on the distribution of features.

[0052] <Effects> According to the image processing device, method, and program of this embodiment, by being provided with the configuration described above, it is possible to improve the accuracy of image processing for determining whether or not a target image contains an image related to a specified object.

[0053] <Variations> In the embodiment described above, the second matching process is performed taking into account the positional relationships and distribution conditions between features according to the present disclosure in order to correct the determination results obtained by the first matching process. However, the matching process taking into account the positional relationships and distribution conditions between features according to the present disclosure is not limited to use in correcting the determination results obtained by other matching processes. Note that in this variation, a description of the configuration common to the embodiment described above will be omitted, and differences will be described.

[0054] 10 is a diagram showing an outline of the functional configuration of an image processing device 1b according to a variation. In the image processing device 1b, a program recorded in the storage device 14 is read into the RAM 13 and executed by the CPU 11, which controls each piece of hardware provided in the image processing device 1b, thereby functioning as an image processing device including a query image acquisition unit 21, a feature extraction unit 22, a match feature identification unit 23, a sample distribution index calculation unit 25, a match distribution index calculation unit 26, an exclusion determination unit 27, and a notification unit 28. That is, in the image processing device 1b according to this embodiment, the object determination unit 24 is omitted from the image processing device 1 shown in FIG. 2.

[0055] Fig. 11 is a flowchart showing the flow of the inspection process related to the variation. Similar to the inspection process described with reference to Fig. 9, the process shown in this flowchart is executed when an instruction to start inspection from a user is received.

[0056] The processes shown in steps S501 to S504 are generally similar to the processes in steps S101 to S104 of the inspection process described with reference to Fig. 9, and the processes shown in steps S505 to S509 are generally similar to the processes in steps S301 to S304 and step S4 of the inspection process described with reference to Fig. 9. That is, in the inspection process shown in this flowchart, steps S105 and S2 are omitted from the inspection process described with reference to Fig. 9.

[0057] <Other Variations> In the embodiment described above, an example has been described in which the sample image is divided into 16 grids arranged in a matrix of 4 rows and 4 columns, and the density of the sample features for each grid is calculated, but the method of dividing the grids is not limited to the example given in this embodiment. For example, a division method in which the sample image is divided into 9 grids arranged in a matrix of 3 rows and 3 columns, or another division method may be used.

[0058] 1. Image processing device

Claims

1. An image processing device comprising: a sample distribution index calculation means for calculating a sample distribution index indicating the variation in a sample image of a plurality of sample features extracted from the sample image relating to a specified object; a match distribution index calculation means for calculating a match distribution index indicating the variation in the sample image of a plurality of match features among the plurality of sample features that match a plurality of query features extracted from a query image to be determined as to whether or not the image relating to the specified object is included; and an exclusion determination means for determining that the query image does not include an image relating to the specified object when a difference between the sample distribution index and the match distribution index exceeds a specified criterion.

2. The image processing device according to claim 1, wherein the sample distribution index calculation means calculates, as the sample distribution index, a standard deviation of the plurality of sample features in the sample image, and the match distribution index calculation means calculates, as the match distribution index, a standard deviation of the plurality of match features in the sample image.

3. The image processing device of claim 2, wherein the sample distribution index calculation means calculates the density of the sample features contained in each grid when the sample image is divided into a plurality of grids, and calculates the standard deviation of the density of the sample features contained in each grid, thereby calculating the sample distribution index; and the match distribution index calculation means calculates the density of the match features contained in each grid when the sample image is divided into a plurality of grids, and calculates the standard deviation of the density of the match features contained in each grid, thereby calculating the match distribution index.

4. The image processing device according to claim 3, wherein the exclusion determination means determines that the query image does not include an image relating to the specified object when a difference between the standard deviation as the sample distribution index and the standard deviation as the match distribution index exceeds a specified threshold.

5. The image processing device of claim 1, further comprising an object determination means for determining whether or not the query image contains an image related to the specified object based on the plurality of match features using a method different from the exclusion determination means, wherein the exclusion determination means determines that the query image does not contain an image related to the specified object when the difference between the sample distribution index and the match distribution index exceeds a specified standard, regardless of the determination result by the object determination means.

6. The image processing device according to claim 5, wherein the object determination means uses a trained machine learning model to determine whether or not the query image contains an image relating to the specified object based on the plurality of match features.

7. The image processing device of claim 1, further comprising a match feature identification means for identifying a plurality of match features among the plurality of sample features that match any of the plurality of query features by determining whether each of the plurality of sample features matches the plurality of query features.

8. The image processing device according to claim 1, further comprising feature extraction means for extracting the plurality of query features from the query image.

9. The image processing apparatus according to claim 8, wherein said feature extraction means further extracts said plurality of sample features from said sample image.

10. An image processing method in which a computer executes the following steps: a sample distribution index calculation step of calculating a sample distribution index indicating the variation in a sample image of a plurality of sample features extracted from the sample image relating to a specified object; a match distribution index calculation step of calculating a match distribution index indicating the variation in the sample image of a plurality of match features that match a plurality of query features extracted from a query image that is to be determined as to whether or not the image relating to the specified object is included among the plurality of sample features; and an exclusion determination step of determining that the query image does not include an image relating to the specified object if a difference between the sample distribution index and the match distribution index exceeds a specified criterion.

11. An image processing program that causes a computer to function as: a sample distribution index calculation means for calculating a sample distribution index indicating the variation in a sample image of a specified object, of a plurality of sample features extracted from the sample image; a match distribution index calculation means for calculating a match distribution index indicating the variation in the sample image, of a plurality of match features that match a plurality of query features extracted from a query image that is to be determined as to whether or not the image related to the specified object is included, among the plurality of sample features; and an exclusion determination means for determining that the query image does not include an image related to the specified object if a difference between the sample distribution index and the match distribution index exceeds a predetermined criterion.

Images