JP2019117466A - Image abnormality determination apparatus, image abnormality determination method and computer program - Google Patents

Abstract

To provide a technique capable of detecting an abnormal area faster and with higher precision.SOLUTION: An image abnormality determination apparatus determines presence or absence of an abnormal area where an abnormality occurs or an area where there is a sign that an abnormality occurs, in an image, on the basis of encoding information generated as a process encoding the image for determination or a result of encoding, and having feature information on a pattern in the image. The image abnormality determination apparatus includes a division area encoding information input unit for acquiring division area encoding information of a specified area among the division area encoding information which is encoding information for an area where the image is divided, a determination information generation unit for generating determination information used for determining whether an abnormal area is included in the specified area, and a determination unit for determining whether the abnormal area is included in the specified area on the basis of the determination information.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an image abnormality determination device, an image abnormality determination method, and a computer program.

  In recent years, a technique for detecting an abnormal part included in a subject at high speed using a visible light image or a sensor image has attracted attention. For example, use cases are efficiency of inspection work based on images of facilities such as solar panels or buildings aerially taken by a drone or identification of a process in which foreign matter is mixed in manufacturing of a semiconductor wafer. In such a use case, it is important to detect an abnormal area at high speed from a large amount of image data captured. In the above-mentioned example, a failure point of equipment, a point where a failure sign is caused due to a foreign substance such as grass grown on a solar panel, or a dust area on a semiconductor wafer is an abnormal area in an image. The method of detecting an abnormal area is, for example, adaptive binarization processing. The adaptive binarization process is a method used for extracting characters included in an image as described in Non-Patent Document 1. In the adaptive binarization process, color information is used. The adaptive binarization process is a process of classifying an image into two for each pixel as 1 when color information is equal to or more than a threshold and as 0 when less than the threshold. The adaptive binarization process can be used to classify an abnormal area and a normal area.

JP, 2016-206372, A

J. Sauvola et. Al., “Adaptive document image binarization,” Pattern Recognition 33 (2), pp. 225-236, 2000.

  However, the prior art has problems with processing speed and detection performance. In the prior art, since color information is used, a pixel value signal is required. In general, image data is encoded. For this reason, decoding processing for image data is required, which requires time. Further, in the abnormal area and the normal area, if there is a difference in shape but the difference in color is small, or if there is a change in color in the normal area, there is a problem that detection omission or erroneous detection occurs.

  In view of the above circumstances, the present invention aims to provide a technique capable of detecting an abnormal area at higher speed and with higher accuracy.

  In one aspect of the present invention, an abnormality occurs in the image based on coding information that is generated as a process of coding an image to be determined or a result of coding and includes feature information of a pattern in the image. It is an image abnormality judging device which judges existence of an abnormal area which is an area where there is an abnormality or an area where there is a sign that an abnormality occurs, and divided area coding information which is coding information for each area where the image is divided. Among them, a divided area coding information input unit that acquires divided area coding information of a specified area, and a determination that generates determination information used to determine whether the abnormal area is included in the specified area According to another aspect of the present invention, there is provided an image abnormality determination apparatus comprising: an information generation unit; and a determination unit that determines whether the specified area includes the abnormal area based on the determination information.

  One aspect of the present invention is the image abnormality determination device described above, wherein the determination information is calculated based on divided region coding information of a predetermined region and divided region coding information of the vicinity of the predetermined region. The determination unit, based on a difference between the divided area coding information of the predetermined area and the statistical value, the presence / absence of an abnormal area in the predetermined area; judge.

  One aspect of the present invention is the above-described image abnormality determination device, wherein the statistical value is any one of an average value, a weighted average value, and a median value.

  One aspect of the present invention is the above-described image abnormality determination device, wherein the determination information is a histogram generated based on coding information related to a partial area or the entire area of the image based on a predetermined condition It is information including a plurality of clusters classified, and the determination unit determines an area classified as a cluster satisfying a predetermined condition among the plurality of clusters as an abnormal area.

  One embodiment of the present invention is the above-described image abnormality determination device, wherein the determination unit is classified into the region classified into the cluster with the smallest number of elements among the plurality of clusters or the cluster with the largest number of elements. One of the two regions is determined as an abnormal region.

  In one aspect of the present invention, an abnormality occurs in the image based on coding information that is generated as a process of coding an image to be determined or a result of coding and includes feature information of a pattern in the image. It is an abnormality determination method executed by an image abnormality determination apparatus that determines the presence or absence of an abnormal area that is an area that is present or an area where there is a sign that an abnormality will occur, wherein the image abnormality determination apparatus A division area coding information input step of acquiring division area coding information of a specified area among the division area coding information which is the coding information of A determination information generation step of generating determination information used to determine whether or not an area is included; and the image abnormality determination apparatus determines the abnormality in the designated area based on the determination information. A determining step of determining whether or not include frequency, an image abnormality determination method having.

  One aspect of the present invention is a computer program for causing a computer to function as the above-described image abnormality determination device.

  According to the present invention, it is possible to provide a technology capable of detecting an abnormal area faster and with higher precision.

It is a figure which shows an example of the said block used as determination object, and the surrounding area of the said block. FIG. 2 is a functional block diagram showing an example of a functional configuration of the image abnormality determination device 100 in the first embodiment. It is a flowchart which shows the example of the flow of the process of an image abnormality determination in 1st Embodiment. It is a figure which shows an example of the histogram 2 with respect to all the code amount in an object image. FIG. 16 is a functional block diagram showing an example of a functional configuration of the image abnormality determination device 200 in the second embodiment. It is a flowchart which shows the example of the flow of the process of an image abnormality determination in 2nd Embodiment.

  Hereinafter, an image abnormality determination device, an image abnormality determination method, and a computer program according to an embodiment of the present invention will be described with reference to the drawings.

First Embodiment
FIG. 1 is a diagram illustrating an example of the block to be determined and a peripheral area of the block. The determination target 1 in FIG. 1 is an image showing a portion of the semiconductor wafer. In FIG. 1, a situation is shown in which specific matter (for example, dust on a semiconductor wafer) is mixed in the block. When the image of the target semiconductor wafer includes a flat semiconductor wafer (normal region) and dust having a complicated shape (abnormal region), the abnormal region is represented as a difference in pattern complexity There is. Although the difference in color is small between the abnormal area and the normal area, the difference in complexity is large. The abnormal area is, for example, an area where an abnormality has occurred or an area where there is a sign that an abnormality occurs. The abnormality may be, for example, a block mixed with foreign matter such as dust or grass. The indication may be, for example, a situation in which an abnormality is predicted in the future by a foreign matter such as dust or grass. Therefore, when it is desired to determine dust on the semiconductor wafer shown in FIG. 1, the image abnormality determining apparatus can determine the presence or absence of an abnormal area by calculating the difference in complexity on the image in the normal area and the abnormal area. .

  The determination target 1 is divided into a plurality of blocks. A block is a processing unit of encoding or a set of them. The minimum processing unit is, for example, 8 × 8 pixels in the case of JPEG (Joint Photographic Experts Group), H.264. In the case of H.264 / AVC (Advanced Video Coding), macroblocks (16 × 16 pixels), In the case of 265 / HEVC (High Efficiency Video Coding), CU (Coding Unit) (8 × 8, 16 × 16, 32 × 32, 64 × 64). Also, a block formed by integrating adjacent blocks may be used as a processing unit. For example, four 8 × 8 CUs may be combined to form a 16 × 16 sized block as a processing unit. Images are encoded block by block. A block is an aspect of a region.

  According to FIG. 1, the determination target 1 is represented by 25 blocks of 5 × 5. A frame 11 surrounded by a dotted line represents the coordinates of the horizontal axis of the block. For example, X represents the block located third from the left in FIG. A frame 12 surrounded by a dotted line represents coordinates of the vertical axis of the block. Y represents the third block from the top of FIG. A block including a region surrounded by a thick frame 13 represents a block including a waveguide superimposed on the base of the semiconductor wafer. A block 14 located at X and Y in FIG. 1 represents the block of the determination target 1. The position of each block of the determination target 1 is represented by coordinates (x, y). For example, the block is represented by coordinates (X, Y). Block 14 contains dust 15. The upper left numerical value of the block is the code amount of each block. That is, the numeral 19 included in the frame 16 represents the code amount of the block located at the coordinate (X + 2, Y + 1).

  FIG. 2 is a functional block diagram showing an example of a functional configuration of the image abnormality determination device 100 in the first embodiment. The image abnormality determination device 100 determines an abnormal area based on encoding information calculated during encoding processing of an image or encoding information of encoded data after encoding. Specifically, the image abnormality determination device 100 determines an abnormal area using information in which the complexity of the pattern in the image is reflected as a numerical value among the encoding information. The image abnormality determination device 100 determines each encoding unit block or set of blocks of encoded data. The coding information is information generated as a process of coding an image or a result of coding. The coding information includes feature information of the pattern in the image to be coded. The coding information is information in which the color property in the image is reflected. The feature information included in the encoding information is, for example, the number of significant coefficients, the number of bins, or the code amount. The image abnormality determination apparatus 100 acquires the numerical value of the coding information for each block serving as a unit of coding such as 8 × 8 pixels. When acquiring the encoded information from the encoded data, the image abnormality determination device 100 can acquire the encoded data without completely decoding the encoded data.

  Based on the difference between the coding information of the block and the coding information of the peripheral blocks of the block, the image abnormality determination device 100 determines whether the complexity of the pattern is significantly different between the block and the peripheral blocks. judge. In addition, the image abnormality determination apparatus 100 determines without being based on the information of the color contained in encoding information. Therefore, the image abnormality determination apparatus 100 does not depend on the determination of the change in color or the magnitude of the difference in color with the peripheral block. The block may be designated from the outside.

  The image abnormality determination apparatus 100 includes a processor such as a CPU (Central Processing Unit) connected via a bus, a memory, an auxiliary storage device, and the like, and executes the image abnormality determination program to calculate the calculation method input unit 101 and the calculation range input unit 102, divided area coding information input unit 103, calculation range code amount input unit 104, reference value calculation unit 105, block code amount input unit 106, absolute value difference calculation unit 107, difference threshold value input unit 108, and difference It functions as an apparatus provided with the threshold value comparison unit 109. Even if all or some of the functions of the image abnormality determination apparatus 100 are realized using hardware such as an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA). Good. The image abnormality determination program may be recorded on a computer readable recording medium. The computer readable recording medium is, for example, a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, or a storage device such as a hard disk built in a computer system. The image abnormality determination program may be transmitted via a telecommunication line.

  The calculation method input unit 101 is configured using an input device such as a touch panel, a mouse, and a keyboard. The calculation method input unit 101 may be an interface for connecting the input device to the image abnormality determination device 100. In this case, the calculation method input unit 101 generates input data (for example, instruction information indicating an instruction to the image abnormality determination device 100) from an input signal input to the input device, and inputs the generated data to the image abnormality determination device 100. The same applies to the calculation range input unit 102, the divided area coding information input unit 103, the calculation range code amount input unit 104, the block code amount input unit 106, and the difference threshold value input unit 108, which are other input units. Configured

  The calculation method input unit 101 receives information specifying the calculation method of the reference value from the outside. The reference value is a numerical value to be compared to determine whether the block is an abnormal area. The reference value is calculated by the reference value calculation unit 105 based on the code amount information of the block and the neighboring blocks. As a calculation method, a method of calculating an average value, a weighted average value, a median value or the like may be used. The calculation method input unit 101 receives information on which at least one of the calculation methods is selected. The calculation method input unit 101 outputs the received selection result of the calculation method to the reference value calculation unit 105.

  The calculation range input unit 102 receives information specifying the calculation range of the reference value. The calculation range is information for specifying how many blocks from the current block to the surrounding blocks are to be regarded as surrounding blocks. For example, when 2 is designated as the calculation range, the peripheral block is defined as a range of ± 2 blocks in height and width from the block. Therefore, up to 5 × 5 blocks around the block are processed as peripheral blocks. The calculation range input unit 102 outputs the calculation range to the calculation range code amount input unit 104.

  The divided area coding information input unit 103 receives the divided area coding information and an address indicating the location on the image of the block. The divided area coding information is coding information for each block into which the image to be determined is divided. The divided area coding information input unit 103 outputs the divided area coding information to the calculation range code amount input unit 104 or the block code amount input unit 106.

  The calculation range code amount input unit 104 specifies divided area coding information of the block based on the address of the block. The calculation range code amount input unit 104 acquires divided area coding information of the block and peripheral blocks included in the calculation range from the block. The calculation range code amount input unit 104 calculates a code amount from the acquired divided area coding information. For example, the code amount of coordinates (x, y) is represented by p (x, y). The calculation range code amount input unit 104 outputs the calculated code amount to the reference value calculation unit 105.

  The reference value calculation unit 105 calculates the reference value S of the block based on the calculation method designation result received from the calculation method input unit 101 and the code amount calculated by the calculation range code amount input unit 104. For example, when the reference value calculation unit 105 receives a calculation method of the average value as a calculation method, a formula of the reference value S is represented by the following formula (1). The reference value calculation unit 105 is an aspect of the determination information generation unit. The determination information generation unit generates determination information used to determine whether an abnormal area is included in the designated area. The determination information is information including divided area coding information of a predetermined area and a statistical value calculated based on the divided area coding information in the vicinity of the predetermined area. The reference value is an aspect of the statistical value.

  When the reference value calculation unit 105 receives a method of calculating a weighted average value as a calculation method, a formula of the reference value S is expressed by the following equation (2). Note that w may be a count value based on Gaussian distribution.

  When the reference value calculation unit 105 receives the calculation method of the median as the calculation method, the equation of the reference value S is expressed by the following equation (3). Z represents a set of peripheral blocks excluding the block. For example, when 2 is received as the calculation range, Z represents a set of 24 blocks excluding the block.

  In an image in which outliers that are not abnormal regions exist, it is desirable to calculate the weighted average value after expanding the median or the calculation range. In this case, the reference value calculation unit 105 can perform processing faster by calculating the median.

  The block code amount input unit 106 specifies divided area coding information of the block based on the address of the block. The block code amount input unit 106 calculates the code amount p (X, Y) from the encoded data of the block. The block code amount input unit 106 outputs the calculated code amount p (X, Y) to the absolute value difference calculation unit 107.

  The absolute value difference calculation unit 107 calculates an absolute value difference value based on the reference value S output from the reference value calculation unit 105 and the code amount of the block output from the block code amount input unit 106. The absolute value difference value is an absolute value of a difference value obtained by subtracting the reference value S from the code amount p (X, Y). The absolute value difference calculation unit 107 outputs the absolute value difference value to the difference threshold value comparison unit 109.

  The difference threshold input unit 108 receives the difference threshold T. The difference threshold value input unit 108 outputs the difference threshold value T to the difference threshold value comparison unit 109. The difference threshold value T is a value for determining whether the block includes an abnormal area. As the difference threshold value T, different values may be used depending on whether an area in which an abnormality occurs as an abnormal area or an area in which a sign of an abnormality occurs is determined.

  The difference threshold value comparison unit 109 compares the absolute value difference value received from the absolute value difference calculation unit 107 with the difference threshold value T received from the difference threshold value input unit 108 so that the block is an abnormal area. It is judged whether or not it contains. The difference threshold value comparison unit 109 makes the determination based on, for example, Formula (4). The difference threshold comparison unit 109 outputs the determination result. The absolute value difference calculation unit 107 and the difference threshold comparison unit 109 are an aspect of the determination unit. The determination unit determines, based on the determination information, whether or not the specified area (for example, the block) includes an abnormal area.

  FIG. 3 is a flowchart showing an example of the process flow of image abnormality determination in the first embodiment. The calculation method input unit 101 receives an instruction to specify the calculation method of the reference value (step S101). The calculation range input unit 102 receives information specifying the calculation range of the reference value (step S102). The divided area coding information input unit 103 receives the divided area coding information of the image to be determined and the address indicating the location on the image of the block (step S103). The calculation range code amount input unit 104 calculates the code of the block and the peripheral blocks of the block (step S104).

  The reference value calculation unit 105 calculates the reference value of the block based on the specification result of the calculation method and the calculated code amount (step S105). The block code amount input unit 106 calculates the code amount of the block (step S106). The absolute value difference calculation unit 107 calculates an absolute value difference value by subtracting the reference value from the value of the block code amount (step S107). The difference threshold input unit 108 receives a difference threshold (step S108).

  The difference threshold comparison unit 109 determines whether the absolute value difference value is equal to or more than the difference threshold (step S109). If the absolute value difference value is not greater than or equal to the difference threshold value (step S110: NO), the difference threshold value comparing unit 109 determines that the block does not include an abnormal area (step S110). If the absolute value difference value is equal to or greater than the difference threshold value (step S110: YES), the difference threshold value comparing unit 109 determines that the block includes an abnormal area (step S111). The difference threshold comparison unit 109 outputs the determination result (step S112).

In the image abnormality determination apparatus 100 configured as described above, the reference value calculation unit 105 calculates the reference value of the block. The absolute value difference calculation unit 107 calculates an absolute value difference value based on the reference value and the value of the block code amount. The difference threshold value comparison unit 109 compares the absolute value difference value with the difference threshold value to determine whether the block includes an abnormal area. Therefore, the image abnormality determination device 100 can determine the abnormal area based on the value of the code amount without being influenced by the difference relating to the change in color. That is, even when the color difference is small between the normal area and the abnormal area or when the color change is also found in the normal area, it is possible to determine the presence or absence of the abnormal area and to determine the abnormal area with high accuracy. become. In addition, since the image abnormality determination device 100 does not need to decode all the encoded data to obtain the encoding information, the abnormality region can be determined at higher speed.
Further, the image abnormality determination device 100 is an image in which the normal area becomes a complicated pattern and the abnormal area becomes a flat space as in the case where it is desired to grasp the dismantled part of the building in the aerial image of the urban area. It can be determined based on the difference in complexity. Therefore, even if the pattern change in the normal area is large, it is possible to determine the abnormal area.
Furthermore, as in the case where it is desired to grasp foreign matter such as grass that has grown on the solar panel, the image abnormality determination device 100 makes the normal area a flat pattern of the solar panel and the abnormal area is complicated by foreign matter such as grass. Even an image to be a pattern can be determined based on the difference in complexity. Therefore, when the image abnormality determination device 100 targets a solar panel, it is possible to detect not only failure but also a sign of failure.

Second Embodiment
FIG. 4 is a diagram showing an example of the histogram 2 for all code amounts in the target image. In FIG. 4, data having two peaks in total, one in each of the small code area and the large code area, will be described as an example. The histogram 2 includes a first cluster 21 and a second cluster 22. The first cluster 21 represents an area with a small code amount. The second cluster 22 represents an area with a large code amount. Each block is classified into the first cluster when the code amount is smaller than the boundary value 23. Each block is classified into a second cluster when the code amount is larger than the boundary value 23. The first cluster has a center of gravity 24. The second cluster 22 has a center of gravity 25. In FIG. 4, among the clusters represented by the histogram 2, a block included in the second cluster 22 which is a cluster with the smallest number of elements is output as an abnormal area. This is because the semiconductor wafer or the solar panel is the number of blocks related to the abnormal area such as dust or grass when the dust is on the semiconductor wafer or the grass grows on the solar panel. It is assumed that the number is smaller than the number of blocks related to the normal area to be That is, when the code amount of each area obtained by dividing such an image is clustered, it is assumed that an abnormal area such as dust or grass is included in a cluster with the smallest number of elements.

  FIG. 5 is a functional block diagram showing an example of a functional configuration of the image abnormality determination device 200 in the second embodiment. The image abnormality determination device 200 determines the presence or absence of an abnormal area based on all the coding information in the specified target image. The image abnormality determination device 200 includes a processor such as a CPU connected via a bus, a memory, an auxiliary storage device, and the like, and executes the image abnormality determination program to execute the divided area coding information input unit 201, the code amount input unit 202, It functions as an apparatus including a parameter input unit 203, a boundary value calculation unit 204, a clustering unit 205, a distance between gravity centers calculation unit 206, a threshold value input unit 207, a threshold value comparison unit 208, and a cluster element number counting unit 209. Note that all or part of the functions of the image abnormality determination device 200 may be realized using hardware such as an ASIC, a PLD, or an FPGA.

  The divided area coding information input unit 201 is configured using an input device such as a touch panel, a mouse, and a keyboard. The divided area coding information input unit 201 may be an interface for connecting the input device to the image abnormality determination apparatus 200. In this case, the divided area coding information input unit 201 generates input data (for example, instruction information indicating an instruction to the image abnormality determining apparatus 200) from the input signal input in the input device, and inputs the generated data to the image abnormality determining apparatus 200. Do. The same applies to the code amount input unit 202, the parameter input unit 203, and the threshold value input unit 207, which are other input units.

  The divided area coding information input unit 201 receives divided area coding information of a divided image to be determined. The divided area coding information input unit 201 outputs the received divided area coding information to the code amount input unit 202. The code amount input unit 202 calculates the code amount from the divided area coding information. The code amount input unit 202 outputs the calculated code amount of each block to the boundary value calculation unit 204 and the clustering unit 205 as a code amount data group. The code amount data group represents the entire code amount of the image to be searched.

  The parameter input unit 203 receives a parameter. The parameters differ depending on the calculation method used by the boundary value calculation unit 204 to calculate the boundary value. The calculation method of the boundary value is, for example, the k-means method capable of classification into a designated number of clusters. When the calculation method is the k-means method, the parameter is the number of clusters. In the case of this embodiment, 2 is designated as the number of clusters. Also, the maximum number of loops and the initial value of each cluster are accepted as parameters. In the present embodiment, 2 is designated as the number of clusters, but a number of 2 or more may be designated.

  The boundary value calculation unit 204 calculates boundary values for classifying the code amount data group received from the code amount input unit 202 into the number of clusters according to the designated parameter. For example, when 2 is given as the number of clusters, the boundary value calculation unit 204 analyzes the distribution of the code amount data group, and calculates the boundary value for classifying into two clusters. The boundary value calculation unit 204 outputs the calculated boundary value to the clustering unit 205. The boundary value calculation unit 204 may use, for example, the k-means method as a method of calculating the boundary value. In this case, when 2 is designated as the number of clusters, the boundary value calculation unit 204 executes the one-dimensional k-means method. Further, the boundary value calculation unit 204 may narrow down the range of code amounts to be the population of the k-means method in order to exclude outliers of the target code amount data group. The range of the amount of code to be narrowed down is accepted as a parameter by the parameter input unit 203. The boundary value calculation unit 204 may calculate the boundary value by a method other than the k-means method. For example, when it is assumed that two peaks exist in the histogram and one local minimum exists between the two peaks, the boundary value calculation unit 204 approximates the histogram with a polynomial function and numerically analyzes the local minimum. The boundary value may be calculated by calculating according to Hereinafter, the number of clusters will be described as 2. However, even if the number of clusters is 2 or more, the abnormal area can be determined.

  The clustering unit 205 generates a code amount data group based on the code amount data group relating to a partial region or the entire region of the image received from the code amount input unit 202 and the boundary value received from the boundary value calculation unit 204. Classify into two clusters. The clustering unit 205 classifies a block whose code amount is smaller than the boundary value into a first cluster. The clustering unit 205 classifies a block whose code amount is larger than the boundary value into a second cluster. The clustering unit 205 outputs the classification result to the distance between centers of gravity calculation unit 206. The clustering unit 205 is an aspect of the determination information generation unit. The first cluster or the second cluster is an aspect of the determination information.

  The inter-centroid distance calculation unit 206 calculates the centers of gravity of the code amounts included in the first cluster and the second cluster. The distance between centers of gravity calculation unit 206 calculates the value D of the distance between centers of gravity based on the calculated centers of gravity. The inter-centroid distance calculation unit 206 outputs the calculated inter-centroid distance value D to the threshold value comparison unit 208.

  The threshold input unit 207 receives the inter-centroid distance threshold U used for comparison with the value D of the inter-centre distance. The inter-centroid distance threshold U is used to determine whether an abnormal area is included in at least a part of the entire area of the image to be determined. The threshold input unit 207 outputs the inter-centroid distance threshold U to the threshold comparison unit 208. The inter-centroid distance threshold value U may be a different value depending on whether an area in which an abnormality occurs as an abnormal area or an area in which a sign of an abnormality occurs is determined.

  The threshold comparison unit 208 makes a determination by comparing the value D of the distance between the centers of gravity received from the distance between the centers of gravity calculation unit 206 with the threshold value of the distance between the centers of gravity U received from the threshold value input unit 207. It is determined whether an abnormal area is included in at least a part of the entire area of the target image. The threshold comparison unit 208 makes the determination based on, for example, the equation (5). The threshold comparison unit 208 outputs the determination result. The inter-centroid distance calculation unit 206 and the threshold comparison unit 208 are an aspect of the determination unit.

  When the value D of the distance between the centers of gravity is equal to or less than the threshold of the distance between the centers of gravity, the threshold comparison unit 208 determines that there is no difference enough to be determined as a peculiar property and no abnormal area is included in the entire area. Output the judgment result. When the value D of the distance between the centers of gravity is larger than the threshold of the distance between the centers of gravity, the threshold comparison unit 208 determines that there is a difference enough to be determined as a peculiar property, and counts the code amount data group as the cluster element number. Output to the part 209.

  The cluster element number counting unit 209 counts the number of elements of code amount data for each of the first cluster and the second cluster. As a result of counting, the cluster element number counting unit 209 determines that a cluster with a small number of elements is a cluster in which an abnormal area is present, and a cluster with a large number of elements is a cluster in which no abnormal area is present. The cluster element number counting unit 209 outputs an area belonging to a cluster with a smaller number of elements as an abnormal area as a determination result.

  FIG. 6 is a flowchart showing an example of the process flow of the image abnormality determination in the second embodiment. In this flowchart, the k-means method is used as a method of calculating the boundary value, and the number of clusters is described as two. The divided area coding information input unit 201 receives divided area coding information of the target divided image. The code amount input unit 202 calculates a code amount data group based on the received coding information (step S201). The parameter input unit 203 receives the value of the number of clusters as a parameter (step S202). In the flowchart, the value of the number of clusters accepted by the parameter input unit 203 is two. The boundary value calculation unit 204 calculates the boundary value based on the calculated code amount data group and the received parameter (step S203).

  The clustering unit 205 classifies the code amount data group into two clusters based on the code amount data group and the boundary value (step S204). Specifically, the clustering unit 205 classifies a block whose code amount is smaller than the boundary value into a first cluster. The clustering unit 205 classifies a block whose code amount is larger than the boundary value into a second cluster. The distance between centers of gravity calculation unit 206 calculates the center of gravity of the code amount data included in the first cluster and the second cluster (step S205). The distance between centers of gravity calculation unit 206 calculates the value of the distance between centers of gravity based on the calculated centers of gravity (step S206).

  The threshold input unit 207 receives an inter-centroid distance threshold used for comparison with the value of the inter-centre distance (step S207). The threshold comparison unit 208 determines whether the value of the distance between the centers of gravity is equal to or less than the threshold of the distance between the centers of gravity (step S208). If the value of the distance between the centers of gravity is equal to or less than the threshold of the distance between the centers of gravity (step S208: YES), the threshold comparison unit 208 determines that no abnormal area is included in the entire area (step S209). The value comparison unit 208 outputs the determination result (step S212).

  When the value of the distance between the centers of gravity is not equal to or less than the threshold of the distance between the centers of gravity (step S208: NO), the cluster element number counting unit 209 counts the number of elements of code amount data for each of the first cluster and the second cluster ( Step S210). The cluster element number counting unit 209 determines the cluster with the smaller number of elements as the cluster with an abnormal area, and the cluster with the larger number of elements as a cluster with no abnormal area based on the counted result (step S211). ). The cluster element number counting unit 209 outputs the determination result (step S212).

  The image abnormality determination apparatus 200 configured as described above calculates the code amount of the image to be determined by the code amount input unit 202 as a code amount data group. The boundary value calculation unit 204 calculates boundary values used to classify clusters. The clustering unit 205 classifies the code amount data group into a plurality of clusters according to the calculated boundary value. The threshold comparison unit 208 determines whether or not an abnormal area is included in the image according to the value of the distance between the centers of gravity of clusters. Therefore, the image abnormality determination device 200 can determine the abnormal area without being affected by the difference related to the change in color. That is, the image abnormality determination apparatus 200 can determine even in the case where there is a small color difference between the normal area and the abnormal area or when the color change is also found in the normal area, the abnormality area with high accuracy. Can be determined. In addition, since the image abnormality determination device 100 does not need to decode all the encoded data to obtain the encoding information, the abnormality region can be determined at higher speed.

  (Modification of First Embodiment and Second Embodiment)

  When the image abnormality determination device has two or more color components of the image to be encoded, the value of the code amount to be processed may process any one component, or the total value of two or more components. May be processed. For example, when the color component is three components of YUV, the image abnormality determination device may process any one of Y, U, or V, or may process the total value of YUV.

  The image abnormality judging device may regard the code amount of a block as a processing unit of encoding as p (x, y), and regards the code amount of a block in which blocks are aggregated as p (x, y) It is also good. When the code amount of a block in which blocks are assembled is regarded as p (x, y), the image abnormality determining apparatus may set a value obtained by totaling the code amounts of the blocks included in the set as p (x, y).

  In the present embodiment, the code amount is taken as the coding information and described, but coding information other than the code amount may be used. For example, the image abnormality determination device may perform the same processing flow using coding information other than the code amount. For example, as coding information other than the code amount, the value of the number of bins before applying entropy coding (bit string before applying CABAC in the case of HEVC), significant coefficient that is nonzero after orthogonal transformation or quantization Information calculated during encoding of the image, such as the number of pixels, or information calculated during decoding of encoded data may be used. Even in the case where coding information other than the code amount is used, the abnormal area can be determined more accurately and at higher speed. In addition, the image abnormality determination device is an image of a cluster even if the normal area is a complicated pattern and the abnormal area is a flat space as in the case where it is desired to grasp a dismantled part of a building in an aerial image of a city area. It can be determined because it is determined by the number of elements. Therefore, even if the pattern change in the normal area is large, it is possible to determine the abnormal area.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design and the like within the scope of the present invention.

  The present invention is applicable to an apparatus for detecting an abnormal area included in a visible light image, a sensor image or the like.

100: Image abnormality determining apparatus, 101: Calculation method input unit, 102: Calculation range input unit, 103: Divisional region coding information input unit, 104: Calculation range code amount input unit, 105: Reference value calculation unit, 106: concerned Block code amount input unit, 107 ... absolute value difference calculation unit, 108 ... difference threshold input unit, 109 ... difference threshold comparison unit, 200 ... image abnormality judging device, 201 ... divided area coding information input unit, 202 ... Code amount input unit, 203 ... Parameter input unit, 204 ... Boundary value calculation unit, 205 ... Clustering unit, 206 ... Center of gravity distance calculation unit, 207 ... Threshold value input unit, 208 ... Threshold value comparison unit, 209 ... Cluster element number counting unit

Claims (7)

A region or abnormality in which an abnormality occurs in the image is generated based on the encoding information including the feature information of the pattern in the image, generated as a process of encoding or a result of encoding the image to be determined. An image abnormality determining apparatus that determines the presence or absence of an abnormal area that is an area where there is a sign that occurs.
A divided area coding information input unit for acquiring divided area coding information of a designated area among divided area coding information which is coding information for each area where the image is divided;
A determination information generation unit that generates determination information used to determine whether or not the abnormal region is included in the designated region;
A determination unit that determines whether the abnormal area is included in the designated area based on the determination information;
An image abnormality determination device comprising: The determination information is information including divided area coding information of a predetermined area, and a statistical value calculated based on divided area coding information in the vicinity of the predetermined area.
The determination unit determines the presence or absence of an abnormal area in the predetermined area based on the difference between the divided area coding information of the predetermined area and the statistical value.
The image abnormality determination device according to claim 1. The statistical value is any one of an average value, a weighted average value or a median value.
The image abnormality determination device according to claim 2. The determination information is information including a plurality of clusters in which a histogram generated based on coding information on a partial area or the entire area of the image is classified based on a predetermined condition.
The determination unit determines an area classified as a cluster that satisfies a predetermined condition among the plurality of clusters as an abnormal area.
The image abnormality determination device according to claim 1. The determination unit determines, of the plurality of clusters, one of an area classified as a cluster having the smallest number of elements and an area classified as a cluster having the largest number of elements as an abnormal area.
The image abnormality determination device according to claim 4. A region or abnormality in which an abnormality occurs in the image is generated based on the encoding information including the feature information of the pattern in the image, generated as a process of encoding or a result of encoding the image to be determined. The abnormality determination method is performed by the image abnormality determination device that determines the presence or absence of an abnormal area that is an area where there is a sign of occurrence.
A divided area coding information input step of acquiring divided area coding information of a designated area among divided area coding information which is coding information for each area where the image is divided;
A determination information generation step of generating determination information to be used for determining whether the specified area includes the abnormal area, the image abnormality determination apparatus;
A determination step of determining whether or not the specified area includes the abnormal area based on the determination information;
An image abnormality judging method having A computer program for causing a computer to function as the image abnormality determining device according to any one of claims 1 to 5.

Images