Image processing method, image processing apparatus, X-ray diffractometer, and program

Abstract

To allow for separating a desired area from a plane image by a simple method.SOLUTION: An image processing method provided herein involves determining whether each of a plurality of pixels belongs to a cluster or not by performing density-based clustering, at a position corresponding to each pixel constituting a plane image, on a plurality of points distributed on a coordinate plane in the number corresponding to a pixel value of the pixel.SELECTED DRAWING: Figure 6

Description

Technical Field

[0001] The present invention relates to a technique for processing various images.

Background Art

[0002] Techniques for separating a target region from an image have been proposed conventionally. For example, Non-Patent Document 1 discloses a technique for separating a region representing a diffraction image and a region representing noise from a diffraction image. Specifically, a region representing a diffraction image and a region representing noise are separated using a learned model (convolutional neural network).

Prior Art Documents

Non-Patent Documents

[0003]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the technique of Non-Patent Document 1 has a problem that it is laborious because a large amount of teacher data (learning data) regarding a diffraction image obtained by separating a diffraction image and background noise is required to generate a learned model. In view of the above circumstances, an object of the present invention is to separate a desired region from a planar image by a simple method.

Means for Solving the Problems

[0005] In order to solve the above problems, the image processing method of the present invention executes density-based clustering on a plurality of points distributed on a coordinate plane in a number corresponding to the pixel value of each pixel corresponding to each pixel constituting a planar image, thereby identifying whether each of the plurality of pixels belongs to a cluster.

Advantages of the Invention

[0006] According to the image processing method of the present invention, without requiring a large amount of learning data, a portion (desired region) where pixels with high pixel values densely exist can be separated as a cluster. That is, it is possible to separate a desired region from a planar image by a simple method.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Mode for carrying out the invention

[0008] <First Embodiment> FIG. 1 is a block diagram illustrating the configuration of an X-ray diffractometer 50 according to a first embodiment of the present invention. The X-ray diffractometer 50 is an apparatus for analyzing the crystal state (for example, crystal structure and degree of crystallization) of an object by observing X-rays diffracted by the object. As illustrated in FIG. 1, the X-ray diffractometer 50 includes an X-ray detector 100 and a processing device 200.

[0009] The X-ray detector 100 is a device capable of detecting X-rays diffracted by an object. Any known device is adopted for the X-ray detector 100. For example, the X-ray detector 100 includes an irradiation unit 11 and a detection unit 12. The irradiation unit 11 is an irradiation device that irradiates an object with X-rays at an incident angle θ.

[0010] The detection unit 12 is a detection device that detects X-rays diffracted by the object among the X-rays irradiated on the object. In the first embodiment, a two-dimensional detector having a detection surface for detecting X-rays is used as the detection unit 12. The detection surface corresponds to a coordinate plane formed by an axis representing the inclination χ of the crystal plane and an axis representing the diffraction angle 2θ. For example, detection elements capable of detecting X-rays are arranged at each position (coordinate) on the detection surface. Specifically, the detection unit 12 generates a signal S (hereinafter referred to as a “detection signal”) representing the intensity of the X-rays detected at each position on the detection surface. That is, the detection signal S is a signal representing the intensity of the X-rays detected for each combination of the diffraction angle 2θ and the inclination χ of the crystal plane.

[0011] The processing device 200 is a device for analyzing the crystal state of an object from the detection signal S generated by the X-ray detection device 100. The processing device 200 of the first embodiment is realized by a computer system including a control device 21, an input device 22, a display device 23, and a storage device 24. For example, an information terminal such as a personal computer is used as the processing device 200.

[0012] The input device 22 receives operations by the user. For example, an operator operated by the user is used as the input device 22. The display device 23 displays various types of information. For example, a liquid crystal display or the like is exemplified as the display device 23.

[0013] The control device 21 is composed of one or a plurality of processing circuits such as a CPU (Central Processing Unit), and comprehensively controls each element of the processing device 200. The storage device 24 is one or a plurality of memories composed of a known recording medium such as a magnetic recording medium or a semiconductor recording medium, and stores programs executed by the control device 21 and various types of data used by the control device 21.

[0014] FIG. 2 is a block diagram illustrating functions realized by the control device 21 executing a program stored in the storage device 24. As illustrated in FIG. 2, the control device 21 of the first embodiment realizes an image generation unit 211 and a chart creation unit 213.

[0015] The image generation unit 211 generates a diffraction image G (an example of a "plane image") from the detection signal S generated by the X-ray detection device 100. FIG. 3 is an example of the diffraction image G generated by the image generation unit 211. The diffraction image G is composed of a plurality of pixels. The diffraction image G corresponds to the detection surface. That is, the position of each pixel of the diffraction image G corresponds to each coordinate in the coordinate plane spanned by the axis representing the inclination χ of the crystal plane and the axis representing the diffraction angle 2θ. Each pixel is set to a pixel value corresponding to the intensity of the X-ray represented by the detection signal S detected at the position corresponding to the pixel on the detection surface. The pixel value of each pixel is proportional to the intensity of the X-ray. It can also be said that the detection signal S is generated for each of the plurality of pixels constituting the diffraction image G.

[0016] The diffraction image G includes a diffraction pattern representing the crystal state of the object. For example, the diffraction pattern is constituted by locally bright portions (hereinafter referred to as "diffraction lines") L. In FIG. 3, a case where the diffraction pattern is constituted by four arc-shaped diffraction lines L (L1 - L4) is illustrated. The diffraction pattern is a portion where pixels with high pixel values are densely present. On the other hand, the portion other than the diffraction pattern in the diffraction image G is a portion including background noise. The background noise is, for example, stray light or noise caused by the X-ray detection device 100. Therefore, it can also be said that the portion other than the diffraction image G is a portion where pixels with high pixel values are sparse (that is, a portion where pixels with low pixel values are dense).

[0017] The diffraction image G is used for generating a diffraction chart. The diffraction chart is a graph representing the intensity of X-rays for each diffraction angle 2θ. Here, assume a configuration (hereinafter referred to as "Comparative Example 1") for generating a diffraction chart from the entire diffraction image G (specifically, one region specified to include all of the plurality of diffraction lines L). That is, in Comparative Example 1, the diffraction chart is generated from a region including not only the diffraction pattern but also the portion other than the diffraction pattern (that is, the portion representing background noise). Specifically, for the entire diffraction image G, the diffraction chart is created by integrating along a predetermined axis (for example, an axis representing the inclination χ of the crystal plane). FIG. 4 shows the diffraction chart generated according to Comparative Example 1. The diffraction chart in FIG. 4 includes a plurality of peaks (hereinafter referred to as "diffraction peaks") P1 - P4 representing the diffraction pattern and a portion representing background noise. Each diffraction peak P corresponds to one diffraction line L.

[0018] The crystal state of the object is specified from the diffraction peak P (for example, peak position and full width at half maximum). However, when background noise is included, the diffraction peak P cannot be accurately specified. Therefore, there is a problem that the crystal state of the object cannot be estimated with high accuracy. In view of the above circumstances, in the present invention, a diffraction chart is created after removing the influence of background noise from the diffraction image G generated by the image generation unit 211.

[0019] The chart creation unit 213 creates a diffraction chart from the diffraction image G generated by the image generation unit 211. The chart creation unit 213 of the first embodiment includes a coordinate generation unit 311, a classification unit 313, and a processing unit 315. The coordinate generation unit 311 and the classification unit 313 are elements that separate the diffraction image from the portion other than the diffraction image in the diffraction image G. The processing unit 315 is an element that creates a diffraction chart from the separated diffraction image.

[0020] Specifically, the coordinate generation unit 311 generates a coordinate plane corresponding to the diffraction image G (referred to as a "diffraction coordinate plane"). FIG. 5 is an example of the diffraction coordinate plane. The diffraction coordinate plane is a coordinate plane spanned by an axis corresponding to the diffraction angle 2θ and an axis corresponding to the inclination χ of the crystal plane. Specifically, the coordinate generation unit 311 generates the diffraction coordinate plane such that points are distributed at positions corresponding to each pixel in a number according to the pixel value of the pixel (that is, the intensity of the X-ray represented by the detection signal S). For example, for a pixel with a pixel value of 10, 10 points are registered at the position corresponding to the pixel. As understood from the above description, in the diffraction coordinate plane, points are arranged at a density according to the pixel value of each pixel at the position corresponding to each pixel. Note that the diffraction image G and the diffraction coordinate plane are stored in the storage device 24.

[0021] The classification unit 313 identifies whether each of a plurality of pixels constituting the diffraction image G belongs to a cluster. Specifically, for a plurality of points distributed in the diffraction coordinate plane, by performing density-based clustering, it is identified whether each pixel belongs to a cluster. For example, DBSCAN (Density-based spatial clustering of applications with noise) is preferably used as density-based clustering. DBSCAN identifies a region where a plurality of points densely exist as a cluster. The region other than the cluster in the set of a plurality of points is a region where points are sparsely and densely present. That is, it can also be said that DBSCAN is a method of separating a set of a plurality of points into a dense region and a sparse and dense region. Note that it is not necessary to preset the number of clusters in DBSCAN.

[0022] The clustering using DBSCAN in the first embodiment is as follows. For each of a plurality of pixels, a region (hereinafter referred to as a "search range") defined by a predetermined radius (hereinafter referred to as a "search distance") from the point corresponding to the pixel is set. That is, it can also be said that a search range is set for each pixel. Then, the number of points existing within the search region is compared with a predetermined threshold value (hereinafter referred to as a "minimum number of neighboring points"). And each point is classified into one of a core point, a reachable point, and an outlier. Note that the minimum number of neighboring points is the minimum value of the points required to form a dense region.

[0023] Core point: There are points in the search range that exceed the minimum number of neighboring points. Reachable point: The points existing within the search range are less than the minimum number of neighboring points, but there are core points within the search range. Outlier: The points existing within the search range are also less than the minimum number of neighboring points, and there are no core points within the search range.

[0024] Pixels corresponding to points existing within the search range centered on a core point belong to the same cluster (an example of a "cluster" in the present invention) as the pixel corresponding to the core point. Pixels corresponding to reachable points belong to the same cluster (an example of a "cluster" in the present invention) as the pixel corresponding to the core point existing at the reachable point. On the other hand, pixels corresponding to outliers do not belong to a cluster. Through the above processing, classification is repeatedly executed for all points corresponding to all pixels in the diffraction coordinate plane until it is identified whether all pixels belong to a cluster.

[0025] In practice, in DBSCAN, pixels corresponding to outliers are identified as clusters to which only they belong. That is, all pixels are classified into some cluster. Therefore, the cluster to which the pixels corresponding to outliers belong (i.e., the cluster representing background noise) is not the target cluster and is excluded from the processing symmetry. Specifically, a predetermined cluster size (hereinafter referred to as the "minimum cluster size") is set, and clusters with a size smaller than the minimum cluster size among a plurality of clusters in the diffraction coordinate plane (i.e., clusters representing noise) are excluded from the processing symmetry. Note that the cluster size is the number of points belonging to the cluster. As understood from the above description, one or more clusters (exemplifications of the "cluster" in the present invention) formed by a plurality of points existing in a number equal to or greater than the minimum cluster size are extracted. The method for setting the search range, the minimum number of neighboring points (exemplification of the "threshold"), and the minimum cluster size will be described later.

[0026] As understood from the above description, by setting the number of points corresponding to the pixel value (intensity of the detection signal S) for each pixel, it becomes possible to cluster a plurality of pixels based on density.

[0027] FIG. 6 is a diffraction coordinate plane after clustering. The dark-colored portion in FIG. 6 corresponds to the cluster C (cluster with a size equal to or greater than the minimum cluster size). FIG. 6 exemplifies a case where a plurality (4) of clusters C1 - C4 are formed for all pixels of the diffraction image G. Each cluster C in the diffraction image G is a portion where pixels with high pixel values are densely present. That is, each cluster C corresponds to one diffraction line L (i.e., diffraction peak P). The set of a plurality of clusters C1 - C4 formed in the diffraction image G corresponds to the diffraction image.

[0028] On the other hand, in the diffraction image G, the area outside the cluster C (hereinafter referred to as "non-target area") is a part where pixels with high pixel values are sparse (that is, a part with a lower pixel value compared to the cluster C). The light-colored part in FIG. 6 corresponds to the non-target area. The non-target area can also be paraphrased as a part representing background noise. As understood from the above description, the classification unit 313 is an element that separates, from the diffraction image G, a part where pixels with high pixel values are densely present (cluster) and a part where pixels with high pixel values are sparse (non-target area). Therefore, the part representing the diffraction image and the part representing the background noise are separated from the diffraction image G.

[0029] The specific setting methods for the search distance, minimum number of neighboring points, and minimum cluster size used in DBSCAN will be described in detail. For example, setting method 1 and setting method 2 are exemplified.

[0030] <Setting method 1> Setting method 1 can be adopted when the diffraction image in the diffraction image G can be clearly confirmed. For example, it is applied when the intensity difference (difference in pixel values) between the diffraction image and the background noise is sufficiently large.

[0031] (1) Search distance The search distance is set according to the allowable error (tolerance) for the diffraction image G. Note that the tolerance is the allowable error for the diffraction angle 2θ with respect to the entire diffraction image G. The tolerance is appropriately determined by the user according to, for example, the specifications of the X-ray detection device 100 and the measurement conditions. Specifically, the pixel length (number of pixels) corresponding to the tolerance is set as the search distance. The search distance is calculated by "tolerance / resolution per pixel". Assume that the tolerance is 0.5° and the resolution per pixel is 0.0193°. In the above case, 0.5° / 0.0193° = 26 is specified as the search distance.

[0032] (2) Minimum number of neighboring points The minimum proximity points are set according to the pixel values of the diffraction image G. For example, the average value of the pixel values per pixel in the diffraction image G is calculated. Then, the product of the average value per pixel and the area of the search region (search distance × search distance × π) is specified as the minimum proximity points.

[0033] (3) Minimum cluster size (2) A value corresponding to the minimum proximity points specified in (2) is set as the minimum cluster size. Typically, a value equal to the minimum proximity points is set as the minimum cluster size.

[0034] <Setting method 2> Setting method 2 can be adopted when the diffraction image in the diffraction image G is unclear. For example, setting method 2 is adopted when the intensity difference (difference in pixel values) between the diffraction image and the background noise is small.

[0035] In setting method 2, OPTICS (Ordering Points To Identify the Clustering Structure) is executed for a plurality of points distributed on the diffraction coordinate plane. OPTICS is a density-based clustering method similar to DBSCAN. In OPTICS, when various values are set for the search distance and the minimum proximity points used in DBSCAN, it is possible to grasp how clusters are formed. Specifically, in OPTICS, the minimum distance required for each point to become a core point is specified. Also in OPTICS, the search distance and the minimum proximity points are used as in DBSCAN.

[0036] The search distance set in OPTICS is arbitrary. For example, a value similar to the (1) search distance in the above-described <Setting method 1> or a value similar to the size in the axial direction representing the diffraction angle 2θ on the diffraction line of the diffraction image is set as the search distance of OPTICS. However, when not considering the calculation cost, a value similar to the size of the diffraction image or infinity may be used as the search distance. It can also be said that the search distance in OPTICS is the upper limit of the distance considered as the search distance of DBSCAN.

[0037] The minimum number of neighboring points set in OPTICS is the same value as the minimum number of neighboring points planned to be set in DBSCAN. The minimum number of neighboring points in OPTICS can be any value greater than or equal to 2. For example, a value similar to the (2) minimum number of neighboring points in <Setting Method 1> described above or a value of about 1 / 4 to 1 / 2 is set as the minimum number of neighboring points in OPTICS.

[0038] Figure 7 is a graph (Reachability Plot) showing the results of OPTICS for the diffraction coordinate plane. Reachability on the vertical axis represents the minimum distance required to become a core point. On the other hand, cluster-ordering on the horizontal axis represents an identifier (information representing the position of the point) for identifying each point. In the Reachability Plot, there are a plurality of peaks and a valley portion K (hereinafter referred to as the "valley portion") located between the peaks. The valley portion K (the portion with little displacement of Reachability) corresponds to a region where points are dense in the diffraction coordinate plane (i.e., the region corresponding to the cluster of the diffraction image G). Therefore, the same number of valley portions K as the number of diffraction lines L in the diffraction image may exist. In Figure 7, there are four valley portions K1 - K4. On the other hand, the peak portion corresponds to a region where points are sparse and dense in the diffraction coordinate plane (i.e., the non-target region).

[0039] (1) Search distance The search distance is set according to the minimum distance of valley portion K in the Reachability Plot. The search distance is set according to the minimum distance of valley portion K2, which has the largest minimum distance among the plurality of valley portions. Valley portion K2 is identified, for example, by calculating the average value of the minimum distance per point for each valley portion K and comparing the average values across the plurality of valley portions K1-K4. Alternatively, the user may visually identify valley portion K2 from the Reachability Plot. Then, for example, a value slightly exceeding the minimum distance (e.g., the average value per point) of valley portion K2 is set as the search distance. Alternatively, the user may set the search distance by visually confirming the minimum distance in valley portion K2. By setting the search distance according to the minimum distance of valley portion K2 with the largest minimum distance (i.e., the valley portion K2 that is the most difficult to form clusters among the plurality of valley portions K), all diffraction lines L constituting the diffraction image G can be separated. Note that the method for identifying the valley portion K with the largest minimum distance and the method for identifying the search distance using the valley portion K are not limited to the above examples.

[0040] (2) Minimum cluster size The minimum cluster size is set according to the width (length in the horizontal axis direction) of valley portion K in the Reachability Plot. Note that the length in the horizontal axis direction is equal to the number of points. Specifically, the minimum cluster size is set according to the width of valley portion K1, which has the smallest width among the plurality of valley portions K1-K4. For example, the minimum cluster size is set to a value equal to the width of valley portion K1. Note that known image analysis techniques are arbitrarily used for identifying valley portion K in the Reachability Plot.

[0041] As understood from the above description, in setting method 2, the search distance and the minimum cluster size are set according to the results of OPTICS. Note that in setting method 2, the same value as the minimum number of neighbors used in OPTICS is used as the minimum number of neighbors in DBSCAN.

[0042] The user can appropriately select whether to adopt Setting Method 1 or Setting Method 2. Setting Method 1 has the advantage that the search distance, the minimum number of adjacent points, and the minimum cluster size can be easily set. On the other hand, Setting Method 2 has the advantage that appropriate search distance and minimum cluster size can be set even when the intensity difference between the diffraction image and the background noise is small. However, the setting methods for the search distance, the minimum number of adjacent points, and the minimum cluster size are not limited to Setting Method 1 and Setting Method 2. In both Setting Method 1 and Setting Method 2, the minimum cluster size can be arbitrary as long as it is a value equal to or greater than the minimum number of adjacent points.

[0043] Note that the search distance, the minimum number of adjacent points, and the minimum cluster size are set for the classification unit 313 before the execution of clustering. For example, the values input by the user operating the input device 22 are set for the classification unit 313 as the search distance, the minimum number of adjacent points, and the minimum cluster size. Further, the elements for specifying the search distance, the minimum number of adjacent points, and the minimum cluster size may be mounted on the processing device 200 or may be mounted on a device separate from the processing device 200.

[0044] The processing unit 315 generates a diffraction chart. Specifically, the processing unit 315 generates a diffraction chart from the region (set of pixels) corresponding to the cluster C in the diffraction image G. That is, the non-target region in the diffraction image G is not taken into account in the diffraction chart.

[0045] First, the processing unit 315 integrates each of the plurality of clusters C1 - C4 with respect to a desired axis (for example, an axis representing the inclination χ of the crystal plane). FIG. 8 shows the result W of the integration for each cluster C. Note that it is not essential to integrate the entire cluster C. For example, integration may be performed on an arbitrary range set by the user within the cluster C (for example, a range set according to the spread of the cluster or the size of the detection unit 12).

[0046] Next, the processing unit 315 fits the integration result W for each cluster C with a desired function (e.g., Gaussian function or Lorentz function). Then, the fitted curved diffraction peaks P1 - P4 are generated. By analyzing the diffraction peak P, the peak position and the full width at half maximum are specified. And the crystal state of the object can be estimated. As understood from the above description, a plurality of diffraction peaks P1 - P4 corresponding to the plurality of clusters C1 - C4 are generated respectively. Note that any known technique is adopted for the process of generating the diffraction peaks P1 - P4 from the cluster C.

[0047] Then, the processing unit 315 creates a diffraction chart by integrating the plurality of diffraction peaks P1 - P4 after fitting. FIG. 9 is the diffraction chart generated by the processing unit 315. Since only the cluster C in the diffraction image G is the target of integration, a diffraction chart with background noise removed is generated as compared with the diffraction chart in FIG. 4. The diffraction chart created by the processing unit 315 is displayed on the display device 23. Note that in the above description, the case where the number of clusters C is 4 is exemplified, but the number of clusters C can be changed according to the diffraction image.

[0048] FIG. 10 is a flowchart of the process executed by the control device 21 of the first embodiment. The process in FIG. 10 is started, for example, in response to an instruction from a user to the processing device 200. First, the image generation unit 211 generates a diffraction image G from the detection signal S (SA1). The coordinate generation unit 311 generates a diffraction coordinate plane corresponding to the diffraction image G (SA2). Specifically, a diffraction coordinate plane in which points are distributed at a number corresponding to the pixel value of each pixel is generated at the position corresponding to each pixel. Actually, a coordinate database corresponding to the diffraction coordinate plane is stored in the storage device 24. The coordinate database is a data table that registers the number of points for each coordinate.

[0049] The classification unit 313 identifies whether each of a plurality of pixels constituting the diffraction image G belongs to a cluster (SA3). Specifically, for a plurality of points distributed on the diffraction coordinate plane, clustering using DBSCAN is performed to identify whether each pixel belongs to a cluster. Then, one or more clusters are formed in the plurality of pixels constituting the diffraction image G. The search distance, the minimum number of neighboring points, and the minimum cluster size used in DBSCAN are set in advance before performing the clustering. The method for specifying the search distance, the minimum number of neighboring points, and the minimum cluster size is as described above.

[0050] The processing unit 315 integrates each of the plurality of clusters C with respect to a desired axis (SA4). Next, the processing unit 315 fits the result of the integration for each cluster C with a desired function (SA5). That is, a diffraction peak P is generated for each cluster C. Then, the processing unit 315 creates a diffraction chart by integrating each diffraction peak P after fitting (SA6).

[0051] Here, in the configuration of Comparative Example 1, as described above, a diffraction chart (the diffraction chart in FIG. 4) including background noise is generated. Therefore, in the configuration of Comparative Example 1, before specifying the peak position and the full width at half maximum of the diffraction peak P, it is necessary to perform a process of reducing background noise from the diffraction chart. For example, the background noise is reduced by performing a Rebbert analysis on the diffraction chart. However, in the execution of the Rebbert analysis, the user needs to input many parameters based on rules of thumb and the like in order to reduce the background noise. Therefore, there is a problem that it is a heavy burden on the user and it also takes time to create the diffraction chart.

[0052] On the other hand, according to the configuration of the first embodiment, by performing density-based clustering on a plurality of points distributed in the diffraction coordinate plane, it is determined whether each of the plurality of pixels of the diffraction image G belongs to a cluster. Then, by performing integration only on the clusters in the diffraction image G, a diffraction chart with background noise removed is created. Therefore, it is not necessary to remove background noise by the user's input after the diffraction chart is created. That is, compared with the configuration of Comparative Example 1, the burden on the user is reduced, and the diffraction chart can be created in a short time.

[0053] In Comparative Example 1, a diffraction chart is generated by performing fitting (for example, fitting with a Gaussian function) on the integration result including a plurality of diffraction peaks. Therefore, there is a problem that the processing load for creating the diffraction chart is large. On the other hand, according to the configuration of the first embodiment in which fitting is performed for each diffraction peak P, the processing load for creating the diffraction chart is reduced compared with Comparative Example 1.

[0054] In a configuration (hereinafter referred to as "Comparative Example 2") that uses a learned model to separate the diffraction image (cluster C) and background noise (non-target region) from the diffraction image G, it is necessary to prepare a large amount of learning data (teacher data regarding the diffraction image G in which the diffraction image and background noise are separated). Comparative Example 2 is a technique described in Non-Patent Document 1, for example. On the other hand, in the configuration of the first embodiment, a large amount of learning data is not required. That is, compared with Comparative Example 2, it is possible to separate the diffraction image and background noise from the diffraction image G in a simple manner.

[0055] For example, a configuration (hereinafter referred to as "Comparative Example 3") has also been proposed in which DBSAN is executed on a binarized diffraction image (i.e., a diffraction image converted into a two-tone image of white and black) to identify clusters. Comparative Example 3 is a technique described in, for example, "Christian Bodenstein, et al., 'Automatic Object Detection Using DBSCAN for Counting Intoxicated Flies in the FLORIDA Assay'". However, with the configuration of Comparative Example 3, it is not possible to perform clustering taking into account the change in pixel values in the diffraction image. In particular, when creating a diffraction chart, it is necessary to separate diffraction images taking into account the spread of the diffraction peak P (i.e., the change in pixel values within the cluster). In the first embodiment, since the points of the number corresponding to the pixel values of the diffraction image G are distributed on the diffraction image G plane, clusters taking into account the change in pixel values can be formed. That is, a diffraction chart appropriately reflecting the spread of the diffraction peak P can be generated.

[0056] <Second Embodiment> The second embodiment of the present invention will be described. For elements whose actions or functions are the same as those in the first embodiment in each of the embodiments illustrated below, the reference numerals used in the description of the first embodiment are reused, and the detailed description of each is appropriately omitted.

[0057] The expected value of the background noise that can be assumed (the intensity of the detection signal S) can be calculated in advance according to the specifications and measurement conditions of the X-ray detection device 100. In view of the above circumstances, in the second embodiment, a configuration for reducing the influence of background noise before performing clustering on the diffraction image G is illustrated.

[0058] Similar to the first embodiment, the image generation unit 211 generates a diffraction image G from the detection signal S. The coordinate generation unit 311 of the second embodiment distributes, on the diffraction coordinate plane, a number of points corresponding to the value obtained by subtracting the expected value of the background noise that can be assumed for one pixel from the pixel value of each of the plurality of pixels constituting the diffraction image G. That is, pixels with a value below the expected value of the background noise per pixel are not registered on the diffraction coordinate plane. On the other hand, for pixels with a value above the expected value of the background noise per pixel, a number of points corresponding to the value obtained by subtracting the expected value of the background noise from the pixel value of the pixel are registered on the diffraction coordinate plane.

[0059] The expected value of the background noise per pixel is, for example, the product of the detection frequency of noise per pixel [count / pixel / s] and the exposure time [s]. Note that the detection frequency of noise per pixel is referred to from the specification of the X-ray detection device 100. However, the method for calculating the expected value of the background noise is arbitrary. For example, a value obtained by multiplying the product of the detection frequency [count / pixel / s] and the exposure time [s] by a predetermined coefficient may be used as the expected value of the background noise per pixel.

[0060] Then, similar to the first embodiment, the classification unit 313 executes clustering on the plurality of points distributed on the diffraction coordinate plane to identify whether each pixel of the diffraction image G belongs to a cluster. Therefore, the cluster C is specified from the diffraction image G. The processing unit 315 creates a diffraction chart from the cluster C in the diffraction image, similar to the first embodiment.

[0061] In the second embodiment, in particular, since a number of points corresponding to the value obtained by subtracting the expected value of the background noise from the pixel value of each pixel are distributed on the diffraction coordinate plane, it is possible to create a diffraction chart with sufficiently reduced influence of background noise. Since pixels with a value below the expected value of the background noise per pixel are not registered on the diffraction coordinate plane, the processing load for generating the diffraction coordinate plane is also reduced.

[0062] In addition, when the expected value of the background noise is sufficiently small, the configuration of the second embodiment may not be adopted. The case where the expected value of the background noise is sufficiently small means, for example, that the expected value of the background noise is at least one digit smaller than the pixel value of the pixel with the minimum pixel value (excluding 0) among the plurality of pixels of the diffraction image G.

[0063] <Third Embodiment> In the first embodiment, the case where the cluster C separated by the classification unit 313 corresponds to the diffraction peak P was exemplified. However, when the intensity difference between the diffraction image and the background noise is small, as illustrated in FIG. 11, a part of the pixels representing the background noise may be specified as the cluster C3. In consideration of the above circumstances, in the third embodiment, a configuration is exemplified in which a diffraction chart is created by excluding the cluster C3 representing the background noise from among the plurality of clusters C separated by the classification unit 313.

[0064] Hereinafter, a method for identifying whether the cluster C separated by the classification unit 313 is a cluster C representing background noise will be exemplified. For example, by visually checking the diffraction coordinate plane or the diffraction image G after performing clustering by the user, it may be identified whether the cluster C is a cluster C representing background noise.

[0065] Alternatively, the control device 21 may compare the number of points existing in the range corresponding to the cluster C in the diffraction coordinate plane with a predetermined threshold value to identify whether the cluster C is a cluster C representing background noise. When the number of points existing in the range corresponding to the cluster C in the diffraction coordinate plane is less than the predetermined threshold value, the cluster C is identified as background noise. The predetermined threshold value is set according to, for example, the minimum number of neighboring points used in DBSCAN. Specifically, a value exceeding the minimum number of neighboring points is set as the threshold value.

[0066] Note that after integrating the plurality of clusters C1-C3 (including the cluster C3 representing background noise) separated by the classification unit 313, the background noise may be removed. FIG. 12 shows the result of integrating the plurality of clusters C (the three cluster regions shown in FIG. 11) separated by the classification unit 313.

[0067] As illustrated in FIG. 12, the integration result for the plurality of clusters C1-C3 contains a portion representing background noise (the portion surrounded by the ellipse) and each diffraction peak P. The average value of the intensity in the portion representing background noise (hereinafter referred to as "noise average value") is calculated from the integration result of FIG. 12. Then, the background noise separated as a cluster is removed by subtracting the noise average value from each diffraction peak P. FIG. 13 is a diffraction chart after fitting after removing the background noise. As illustrated in FIG. 13, it was confirmed that the background noise was reduced from each diffraction peak P. Note that the method of specifying the portion representing background noise from the integration result for the plurality of clusters C1-C3 is arbitrary. For example, the user may specify it visually, or it may be specified by a known image analysis technique.

[0068] Note that if, as a result of performing clustering by setting the search distance, the minimum number of neighboring points, and the minimum cluster size in the setting method 1, there is a cluster C representing background noise, clustering may be performed again after resetting an appropriate search distance, the minimum number of neighboring points, and the minimum cluster size.

[0069] <Modification Example> Each of the forms exemplified above can be variously modified. Specific modification modes are exemplified below. It is also possible to appropriately combine two or more modes arbitrarily selected from the following examples.

[0070] (1) In each of the above-described embodiments, DBSCAN is used as density-based clustering, but the density-based clustering method is not limited to DBSCAN. For example, HDBSCAN (Hierarchical Density-Based Spatial Clustering) or DENCLUE (Density Based Clustering) may be used. Known clustering techniques can be arbitrarily adopted.

[0071] (2) In each of the above-described embodiments, a configuration for separating from the diffraction image G a portion where pixels with high pixel values are densely present and a portion where pixels with high pixel values are sparsely present is exemplified, but the planar image to be processed is not limited to the diffraction image G. For example, the present invention can be used for various planar images such as an image obtained by imaging a fingerprint used for fingerprint authentication, an X-ray image obtained by irradiating an object with X-rays, or an image obtained by CT (Computed Tomography) or MRI (Magnetic Resonance Imaging). As understood from the above description, any planar image can be the object of the processing of the present invention.

[0072] (3) In each of the above-described embodiments, the functions of the control device 21 of the processing device 200 may be realized by a plurality of devices. For example, the image generation unit 211 and the chart creation unit 213 may be mounted on separate devices. Further, the control device 21 may also realize functions different from those of the image generation unit 211 and the chart creation unit 213. Note that the device equipped with the classification unit 313 is an example of the "image processing device" according to the present invention.

[0073] (4) The functions of the image processing device according to each of the above-described embodiments are realized by the cooperation of a processing circuit such as a CPU and a program as exemplified in each embodiment. The program according to each of the above-described embodiments can be provided in a form stored in a computer-readable recording medium and installed in a computer. The recording medium is, for example, a non-transitory recording medium, and an optical recording medium (optical disk) such as a CD-ROM is a preferred example, but any known form of recording medium such as a semiconductor recording medium or a magnetic recording medium is also included.

[0074] (5) The process performed by the classification unit 313 is also specified as an image processing method for identifying whether each of a plurality of pixels belongs to a cluster by performing density-based clustering on a plurality of points distributed on the coordinate plane at positions corresponding to the pixels constituting the planar image, in accordance with the number corresponding to the pixel value of each pixel.

Explanation of Signs

[0075] 11: Irradiation unit 12: Detection unit 21: Control device 22: Input device 23: Display device 24: Storage device 50: X-ray diffractometer 100: X-ray detector 200: Processing device 211: Image generation unit 213: Chart creation unit 311: Coordinate generation unit 313: Classification unit 315: Processing unit C: Cluster G: Diffraction image L: Diffraction line P: Diffraction peak S: Detection signal

Claims

1. For the coordinate plane of a planar image, register a number of points corresponding to the pixel value of each of a plurality of pixels constituting the planar image at the coordinates corresponding to each of the plurality of pixels, and perform density-based clustering on the plurality of points registered for the coordinate plane to identify whether each of the plurality of pixels belongs to a cluster. In the clustering, DBSCAN is used. For each of the plurality of pixels, identify whether the pixel belongs to the cluster by comparing the number of points existing within a range defined by a predetermined radius centered on the point corresponding to the pixel with a predetermined threshold value. The cluster is formed by a plurality of points existing in a number equal to or greater than the threshold value. Image processing method.

2. The threshold value is set according to the product of the average of the pixel values across the plurality of pixels and the area of the range. The image processing method according to Claim 1.

3. The radius is set according to the allowable error for the planar image. The image processing method according to Claim 1 or Claim 2.

4. The radius and the number of points forming the cluster are set according to the result of OPTICS for the coordinate plane. The image processing method according to Claim 1.

5. For the coordinate plane of a planar image that is a diffraction image representing the intensity distribution of X-rays diffracted by an object, register a number of points corresponding to the pixel value of each of a plurality of pixels constituting the planar image at the coordinates corresponding to each of the plurality of pixels, and perform density-based clustering on the plurality of points registered for the coordinate plane to identify whether each of the plurality of pixels belongs to a cluster. Create a diffraction chart by integrating the regions corresponding to the clusters among the plurality of pixels. Image processing method.

6. For each of the plurality of pixels, register a number of points corresponding to the value obtained by subtracting the expected value of noise that can be assumed for one pixel from the pixel value of the pixel. The image processing method according to Claim 1 or Claim 5.

7. For the coordinate plane of a planar image, it includes a classification unit that registers a number of points corresponding to the pixel value of each of a plurality of pixels constituting the planar image at the coordinates corresponding to each of the plurality of pixels, and performs density-based clustering on the plurality of points registered for the coordinate plane to identify whether each of the plurality of pixels belongs to a cluster. In the clustering, DBSCAN is used. ​ For each of the plurality of pixels, the classification unit compares the number of points existing within a range defined by a predetermined radius centered on the point corresponding to the pixel with a predetermined threshold value, thereby identifying whether the pixel belongs to the cluster. The cluster is formed by a plurality of points existing in a number equal to or greater than the threshold value. An image processing apparatus.

8. Regarding the coordinate plane of a planar image that is a diffraction image representing the intensity distribution of X-rays diffracted by an object, for each of the plurality of pixels constituting the planar image, a number of points corresponding to the pixel value of the pixel are registered at the coordinates corresponding to each of the pixels, and for the plurality of points registered on the coordinate plane, density-based clustering is performed to identify whether each of the plurality of pixels belongs to a cluster. A classification unit A processing unit that creates a diffraction chart by integrating the regions corresponding to the cluster among the plurality of pixels. An image processing apparatus comprising the above.

9. An X-ray detection device that detects X-rays diffracted by an object, An image generation unit that generates a diffraction image representing the intensity distribution of the detected X-rays, Regarding the coordinate plane of the diffraction image, for each of the plurality of pixels constituting the diffraction image, a number of points corresponding to the pixel value of the pixel are registered at the coordinates corresponding to each of the pixels, and for the plurality of points registered on the coordinate plane, density-based clustering is performed to identify whether each of the plurality of pixels belongs to a cluster. A classification unit A processing unit that creates a diffraction chart by integrating the regions corresponding to the cluster among the plurality of pixels. An X-ray diffractometer comprising the above.

10. Function a computer system as a classification unit that, for the coordinate plane of a planar image, registers a number of points corresponding to the pixel value of each pixel at the coordinates corresponding to each of the plurality of pixels constituting the planar image, and performs density-based clustering on the plurality of points registered on the coordinate plane to identify whether each of the plurality of pixels belongs to a cluster. In the clustering, DBSCAN is used. For each of the plurality of pixels, the classification unit compares the number of points existing within a range defined by a predetermined radius centered on the point corresponding to the pixel with a predetermined threshold value, thereby identifying whether the pixel belongs to the cluster. The cluster is formed by a plurality of points existing in a number equal to or greater than the threshold value. A program. For the coordinate plane of a planar image that is a diffraction image representing the intensity distribution of X-rays diffracted by an object, for each coordinate corresponding to each of a plurality of pixels constituting the planar image, register a number of points corresponding to the pixel value of the pixel, and for the plurality of points registered on the coordinate plane, execute density-based clustering to identify whether each of the plurality of pixels belongs to a cluster, and, A processing unit that creates a diffraction chart by integrating the regions corresponding to the clusters among the plurality of pixels A program that causes a computer system to function as.

Images