Processing apparatus, processing method, and program

Abstract

To easily and accurately estimate the composition of each phase in a substance composed of multiple combined phases. [Solution] The processing apparatus 100 performs image processing on a photograph of a cross-section of a material composed of N phases bonded together. This processing includes deriving a frequency distribution of pixel values ​​of the image 410 to be processed, and determining M-1 threshold values ​​Th based on the result of curve fitting of the frequency distribution. j The process of deriving and M-1 thresholds Th j Based on this, a process is performed to generate an image as the first M-value image 420 in which the pixel values ​​of the image to be processed 410 have been converted to M values, and the pixel values ​​of the phase boundary peripheral region in the image to be processed 410 and M-1 thresholds Th j Based on this, the process includes generating a second megaquantized image 430 by correcting the pixel values ​​in the phase boundary peripheral region of the first megaquantized image 420.

Description

[Technical Field]

[0001] This disclosure relates to an apparatus, a processing method, and a program, and is particularly suitable for use in analyzing the composition of a substance composed of multiple combined phases. [Background technology]

[0002] To identify the quality of a substance composed of multiple phases, its composition is analyzed. One example of such a substance is sintered ore. Sintered ore is produced by mixing several types of raw materials, such as powdered iron ore, limestone, and powdered coke, and then sintering these raw materials. It is used as a major raw material in blast furnaces. Sintered ore is a sintered body of the aforementioned multiple types of raw materials, and its interior is porous and consists of multiple microscale phases (hereinafter referred to as "microstructure" as needed). Typical microstructures of sintered ore include hematite (Fe2O3), magnetite (Fe3O4), silicate slag (CaO·SiO2), and calcium ferrite (CaO·Fe2O3). Cross-sections of sintered ore are observed using microscopes such as optical microscopes and electron microscopes (e.g., scanning electron microscopes (SEM)). In addition to the microstructure, voids also appear in the observed image (captured image) of the cross-section of sintered ore. The characteristic features of these phases (e.g., shape and phase fraction) are known to be strongly correlated with the quality of the sintered ore (e.g., reducing properties and strength). Therefore, the rapid and accurate identification of these phases is an extremely important technique in the production of sintered ore.

[0003] One such technology is described in Patent Document 1. In Patent Document 1, first, multiple training data is generated by associating each of multiple regions of a cross-sectional image (captured image) of a cross-section of sintered ore with the structural composition of the composite structure (microstructure) formed by the mineral phases contained in the sintered ore. Using these multiple training data, a learning model is trained on the relationship between multiple cross-sectional images of sintered ore and the structural composition of the composite structure. Based on the relationship between the multiple cross-sectional images and the structural composition of the composite structure learned in this way, a learning model is output that estimates the structural composition of the composite structure of sintered ore from the cross-sectional image of the sintered ore. In Patent Document 1, the structural composition of the composite structure of sintered ore is estimated using a learning model created by performing such supervised learning. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2021-166002 [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] However, the technology described in Patent Document 1 estimates the composition of each phase of a substance composed of multiple combined phases using a learning model. Therefore, if, for example, the conditions for preparing a sample of the substance to be photographed or the conditions for taking cross-sectional images differ between the learning and estimation phases, it is not easy to accurately estimate the composition of each phase of the substance. Therefore, these conditions must be the same between the learning and estimation phases. To do this, a large amount of training data must be prepared. As described above, the technology described in Patent Document 1 does not easily estimate the composition of each phase of a substance composed of multiple combined phases with high accuracy.

[0006] This disclosure has been made in view of the above-mentioned problems, and aims to easily and accurately estimate the composition of each phase of a substance composed of multiple phases combined. [Means for solving the problem]

[0007] The apparatus of the present disclosure comprises: an acquisition unit that acquires an image of a cross-section of a material composed of N phases bonded together, captured by a microscope equipped with an imaging means; and an image processing unit that performs processing including image processing on the captured image, wherein the image processing unit includes: a frequency distribution derivation unit that derives a frequency distribution of pixel values ​​of the image to be processed; a threshold derivation unit that derives M-1 thresholds for converting the image to be processed into an M-value image based on the result of curve fitting the frequency distribution; and an image of the image to be processed with M values ​​obtained by converting the pixel values ​​of the image to be processed into an M-value image based on the M-1 thresholds. The system comprises a first multilevel image generation unit that generates an image as a first M-level image, and a second multilevel image generation unit that identifies regions that form the boundaries of a plurality of phases in the image to be processed, and generates a second M-level image in which the pixel values ​​of the regions that form the boundaries of the plurality of phases in the first M-level image have been corrected based on the pixel values ​​of the identified regions and the M-1 thresholds, wherein N is an integer of 2 or more, M is an integer between 2 and N, and the image to be processed is the captured image or an image that has been preprocessed from the captured image.

[0008] The processing method of the present disclosure comprises an acquisition step of acquiring an image taken of a cross-section of a material composed of N phases bonded together using an optical microscope equipped with an imaging means, and an image processing step of performing image processing on the acquired image, wherein the image processing step comprises a frequency distribution derivation step of deriving a frequency distribution of pixel values ​​of the image to be processed, a threshold derivation step of deriving M-1 thresholds for converting the image to be processed into an M-value image based on the result of curve fitting the frequency distribution, and converting the pixel values ​​of the image to be processed into M-values ​​based on the M-1 thresholds. The process comprises: a first multilevel image generation step which generates an image as a first M-level image; and a second multilevel image generation step which identifies regions that form the boundaries of multiple phases in the image to be processed, and generates an image as a second M-level image in which the pixel values ​​of the regions that form the boundaries of multiple phases in the first M-level image are corrected based on the pixel values ​​of the identified regions and the M-1 thresholds, wherein N is an integer of 2 or more, M is an integer between 2 and N, and the image to be processed is the captured image or an image that has been preprocessed from the captured image.

[0009] The program of this disclosure causes a computer to function as one of the components of the processing unit. [Effects of the Invention]

[0010] According to this disclosure, the composition of each phase of a substance composed of multiple phases can be easily and accurately estimated. [Brief explanation of the drawing]

[0011] [Figure 1] This figure shows a first example of the functional configuration of the processing unit. [Figure 2] This figure shows an example of an image captured without any image compositing and an example of an image captured with image compositing. [Figure 3] This figure shows an example of the frequency distribution of pixel values ​​in an image to be processed. [Figure 4] This diagram illustrates an example of processing applied to an image. [Figure 5]This is a flowchart illustrating the first example of the processing method. [Figure 6] This figure shows a second example of the functional configuration of the processing unit. [Figure 7] This diagram illustrates an example of a process that changes the pixel values ​​of an object surrounded by a void. [Figure 8] This is a flowchart illustrating a second example of the processing method. [Figure 9] This figure shows a third example of the functional configuration of the processing unit. [Figure 10] This figure shows an example of the threshold change screen. [Figure 11] This is a flowchart illustrating a third example of the processing method. [Modes for carrying out the invention]

[0012] Embodiments of this disclosure will be described below with reference to the drawings. Furthermore, the statement that multiple values ​​being compared, such as length, position, size, and spacing, are the same includes not only cases where they are strictly identical, but also cases where they differ to the extent that they do not deviate from the spirit of this disclosure. For example, if multiple values ​​differ within the tolerance range defined at the time of design, those multiple values ​​may be considered the same. Also, for example, if each of multiple values ​​is within the tolerance range, those multiple values ​​may be considered the same. In this case, those multiple values ​​may be measured values. If those multiple values ​​are measured values, the tolerance may be, for example, the maximum tolerance defined in the measuring instrument used to measure those values. Also, those multiple values ​​may be estimated values ​​(calculated values). If those multiple values ​​are estimated values ​​(calculated values), the tolerance may be the error that arises from the calculation of those values ​​being performed by a computer. Also, the tolerance may be the error that arises from the modeling of the actual phenomenon using mathematical formulas.

[0013] (First Embodiment) First, the first embodiment will be described. <Processing device 100> Figure 1 shows an example of the functional configuration of the processing unit 100. The processing unit 100 has, for example, one or more hardware processors such as a CPU (Central Processing Unit), and one or more memory such as RAM (Random Access Memory) and ROM (Read Only Memory) as hardware. The processing unit 100 performs various calculations, for example, by executing one or more programs stored in memory using one or more hardware processors.

[0014] Furthermore, the processing unit 100 is connected to an input device 110 and an output device 120 in a manner that allows communication with the processing unit 100. The communication between the processing unit 100 and the input device 110 and output device 120 may be wired communication or wireless communication. The processing unit 100 may also include the input device 110 and the output device 120.

[0015] Furthermore, the processing unit 100 may be implemented using dedicated hardware such as an ASIC (Application Specific Integrated Circuit).

[0016] The processing apparatus 100 performs processing to estimate the composition of each phase (microstructure) of a substance composed of N combined phases. N is an integer of 2 or more. Note that all types of phases (N types of phases) constituting the substance may be known, or some phases may not be known. For example, if the substance contains impurity phases, the type of such impurity does not need to be specified. However, it is preferable that the types of the M phases to be estimated in the processing apparatus 100 are specified. M is an integer of 2 or more and less and less than or equal to N. As will be described later, the processing apparatus 100 of this embodiment can estimate the composition of each phase even if the number of phases constituting the substance is large. Therefore, M may be an integer of 3 or more. In this case, N is also an integer of 3 or more. In this embodiment, the case in which the substance composed of N combined phases is a sintered ore is given as an example. Therefore, the processing apparatus 100 of this embodiment estimates the composition of multiple phases, including the mineral phase of the sintered ore. Note that there may be one or more vacancies in at least one of the interior and surface of the sintered ore. In this embodiment, the case in which the gas present in the vacancies is treated as one phase is given as an example. In other words, this embodiment exemplifies a case where the M phases to be estimated include gas present in vacancies. In the following description, the gas present in vacancies, which are the M phases to be estimated, will be referred to as vacancies as needed. Furthermore, the substance composed of N phases is not limited to sintered ore. For example, the substance composed of N phases may be coal, coke, lump ore, ironmaking pellets, stainless steel, or TRIP steel (transformation-induced plasticity steel), etc.

[0017] In Figure 1, this embodiment illustrates a case where the processing unit 100 comprises an acquisition unit 101, an image processing unit 102, and an output unit 103. An example of each part of the processing unit 100 in this embodiment is described below.

[0018] <<Acquisition part 101>> The acquisition unit 101 acquires input information that needs to be input to the processing device 100 in order for the processing device 100 to perform processing. In this embodiment, an example is given where the input information includes an image taken of a cross-section of sintered ore with a microscope equipped with an imaging means. In this embodiment, an example is given where the microscope equipped with an imaging means is an optical microscope equipped with an imaging means. However, the microscope is not limited to an optical microscope. For example, the microscope may be an electron microscope (e.g., a scanning electron microscope). In this embodiment, an example is given where the image taken is compressed in a predetermined compression format such as JPEG (Joint Photographic Experts Group) and saved as an image file. The compression format may be lossless or lossy compression. In addition, the image taken does not have to be compressed. For example, the image taken may be saved as an image file in a file format such as TIFF (Tagged Image File Format), BMP (bitmap), PNG (Portable Network Graphics), or GIF (Graphics Interchange Format), and the method of this embodiment can be applied to such images as well. In this embodiment, an example is given where the image taken is a grayscale image.

[0019] A sample is prepared to photograph the cross-section of sintered ore using a microscope equipped with imaging capabilities. To prepare the sample, for example, sintered ore having a size greater than or equal to the size required for the product (e.g., sintered ore with a particle size of 5 mm or larger) is embedded in resin. Then, the surface of the resin in which the sintered ore is embedded is polished. The resin with the surface thus polished is used as the sample. In this case, the polished surface of the sample is the observation surface of the microscope. It is preferable that the imaging conditions (magnification, etc.) be set so that the widest possible area (preferably the entire area) of the cross-section of the sintered ore on the polished surface of the sample is included in a single image. Note that the method of preparing the sample and the method of photographing the cross-section of the material can be realized using known techniques. Therefore, a detailed explanation of these is omitted here.

[0020] In this embodiment, we illustrate a case in which the acquisition unit 101 acquires the captured image obtained as described above from the input device 110. The input device 110 may include, for example, a communication device connected to a communication device connected to an optical microscope via a communication cable. In this case, the input device 110 receives, for example, a captured image transmitted via a communication cable from the communication device connected to the optical microscope. In this case, the acquisition unit 101 acquires the captured image received by the input device 110. Note that the input device 110 and the communication device connected to the optical microscope may communicate wirelessly or via a network. Furthermore, the input device 110 may include a storage medium. In this case, the input device 110 stores the captured image. In this case, the acquisition unit 101 acquires the image by reading it from the input device 110.

[0021] Furthermore, in this embodiment, we illustrate a case where the acquisition unit 101 also acquires input information other than captured images from the input device 110. The input device 110 may be equipped with a user interface. In this case, an operator may input information such as numerical values ​​indicating input information by operating the user interface. In this case, the acquisition unit 101 acquires the input information input in this manner. The input device 110 may also be equipped with a communication device. In this case, the input device 110 may receive input information other than captured images from an external device. In this case, the acquisition unit 101 acquires the input information received by the input device 110 in this manner. The input device 110 may also be equipped with a storage medium in which input information other than captured images is stored. In this case, the acquisition unit 101 acquires the input information by reading it from the storage medium.

[0022] <<Image processing unit 102>> The image processing unit 102 performs processing including image processing on the captured image. In this embodiment, an example is given in which the image processing unit 102 has a preprocessing unit 102a, a frequency distribution derivation unit 102b, a threshold derivation unit 102c, a first multi-level image generation unit 102d, a second multi-level image generation unit 102e, and a feature quantity derivation unit 102f.

[0023] <<<Pre-processing unit 102a>>> The preprocessing unit 102a performs preprocessing on the captured image. Preprocessing includes processes aimed at making the multiple phases displayed in the captured image clearer. Thus, this embodiment illustrates a case in which preprocessing is performed on the captured image. However, it is not always necessary to perform at least one of the preprocessing steps described below. For example, if the difference in the results of the multi-level conversion (M-level conversion) described later is small or nonexistent depending on whether or not preprocessing is performed, or if M phases are clearly displayed in the captured image, then at least one of the preprocessing steps described below may not be performed.

[0024] Generally, sintered ore contains hematite, magnetite, silicate slag, and calcium ferrite. The inventors have found that the brightness in captured images tends to be highest in the order of hematite, magnetite, calcium ferrite, silicate slag, and vacancies (i.e., the brightness of vacancies is the lowest, and the brightness of hematite is the highest). Therefore, these phases can be distinguished by the brightness information obtained from the captured image. As mentioned above, in this embodiment, the case in which the captured image is a grayscale image is given as an example. When the pixel value is represented by 8 bits, the pixel value of a grayscale image is represented by 256 gradation values ​​(brightness values). That is, the pixel value of the captured image represents the brightness. In the processing in the processing apparatus 100, in order to easily extract brightness information (brightness information) from the pixel value of the captured image, if the captured image is a color image, the preprocessing may include a process to convert the captured image to a grayscale image. Thus, the image to be processed may be a captured image represented in grayscale, or a captured image converted to a grayscale image may be used as the image to be processed. Alternatively, color images that have not been converted to grayscale may be used as the images to be processed.

[0025] Furthermore, as mentioned above, this embodiment exemplifies a case where the captured image is compressed in a predetermined compression format such as JPEG (Joint Photographic Experts Group) and saved as an image file. In this case, the captured image may contain noise caused by image compression, such as mosquito noise. To reduce such noise, the preprocessing may include a process to smooth the captured image. The process of smoothing the captured image itself can be implemented using known techniques. In this embodiment, the preprocessing unit 102a exemplifies a case where it smooths the captured image by using a Gaussian filter, which is an example of a smoothing filter. Note that the smoothing filter is not limited to a Gaussian filter. For example, the smoothing filter may be an averaging filter. In this embodiment, the preprocessing unit 102a exemplifies a case where it smooths a captured image represented in grayscale.

[0026] Furthermore, in order to make the boundaries between multiple phases clearer, the preprocessing may include a process to enhance the edges in the captured image, and a process to combine the pixel values ​​of the captured image without edge enhancement with the pixel values ​​of the captured image with edge enhancement. Note that the captured image without edge enhancement is the image before the edge enhancement process is performed. Also, the edges in the captured image correspond to the boundaries (contours) of multiple phases.

[0027] The process of enhancing edges in a captured image is implemented using known techniques. In this embodiment, we illustrate a case where the preprocessor 102a enhances edges in a captured image by using a Sobel filter, which is an example of an edge filter. Note that the edge filter is not limited to a Sobel filter. For example, the edge filter may be a Laplacian filter. If (relatively large) noise is included in the captured image, the process of enhancing the boundaries between multiple phases in the captured image may enhance the noise. Therefore, it is preferable for the preprocessor 102a to perform the edge enhancement process on a smoothed captured image. In this embodiment, we illustrate a case where the preprocessor 102a performs the edge enhancement process on a captured image after smoothing a captured image represented in grayscale.

[0028] The preprocessor 102a then combines the pixel values ​​of the captured image without edge enhancement and the pixel values ​​of the captured image with edge enhancement. To do this, the preprocessor 102a calculates a single pixel value for each pixel in the captured image using the pixel value of the pixel in the captured image without edge enhancement and the pixel value of the pixel corresponding to that pixel in the captured image with edge enhancement. In this case, the preprocessor 102a may calculate the average value of the pixel values ​​of the corresponding pixels in the captured image without edge enhancement and the captured image with edge enhancement. In this embodiment, the case where the average value is an arithmetic mean is given as an example. However, the average value is not limited to an arithmetic mean. The average value may also be a weighted mean. In this case, for example, the weight coefficient for the pixel value of the captured image with edge enhancement may be larger than the weight coefficient for the pixel value of the captured image without edge enhancement. Alternatively, instead of the average value, for example, an additive value may be used. In the following description, the process of combining the pixel values ​​of a captured image without edge enhancement and the pixel values ​​of a captured image with edge enhancement will be referred to as the "combination process" as needed. Furthermore, in this embodiment, the case in which the captured image without edge enhancement is a captured image that has undergone smoothing processing will be used as an example.

[0029] Figure 2 shows an example of an unprocessed image (smoothed image) 210 (Figure 2(a)) and an image that has undergone processing (Figure 220) (Figure 2(b)). In Figures 2(a) and 2(b), by combining the pixel values ​​of the captured image without edge enhancement and the pixel values ​​of the captured image with edge enhancement, the luminance (brightness) of regions 211a and 211b shown in Figure 2(a) is corrected to be as shown in regions 221a and 221b shown in Figure 2(b).

[0030] In this embodiment, we illustrate the case where the captured image that has undergone the synthesis process as described above is the image to be processed. In this case, the image to be processed is the captured image used in the image processing unit 102 for processing after preprocessing. For example, if no preprocessing is performed, the captured image may be the image to be processed. In this case, the processing unit 100 does not need to include the preprocessing unit 102a.

[0031] <<Frequency distribution deriving unit 102b>> The frequency distribution derivation unit 102b derives the frequency distribution of pixel values ​​in the image to be processed. In this embodiment, since the image to be processed is a grayscale image, the pixel values ​​correspond to brightness. Figure 3 shows an example of the frequency distribution 300 of pixel values ​​in the image to be processed. In the following description, the frequency distribution of pixel values ​​in the image to be processed will be abbreviated as frequency distribution as needed. In Figure 3, for the sake of notation and explanation, the frequency distribution 300 is represented by a smooth curve. The frequency distribution may also be represented, for example, by a point cloud on a two-dimensional Cartesian coordinate system with frequency on the y-axis and pixel value on the x-axis.

[0032] <<Threshold Derivation Unit 102c>> The threshold derivation unit 102c derives M - 1 thresholds for converting the processing target image into an M - valued image based on the result of curve - fitting the frequency distribution 300 (with a predetermined function). When the value of M is small (for example, when it is 2 or 3), the threshold derivation unit 102c may curve - fit the frequency distribution 300 with one function (for example, a quadratic function or a quartic function) as the predetermined function. However, when the value of M is large, it may not be easy to curve - fit the frequency distribution 300 quickly and with high precision by using one function. Therefore, in this embodiment, an example is illustrated in which the threshold derivation unit 102c curve - fits the frequency distribution 300 with a distribution function formed by combining M functions each having one upward - convex peak. In this embodiment, an example is illustrated in which the distribution function is represented by a mixture Gaussian distribution. In this case, the distribution function is derived by linearly combining M Gaussian functions. Note that curve - fitting itself can be realized by a known technique such as the least - squares method, so the detailed description thereof is omitted here.

[0033] Note that the mixture Gaussian distribution P(x) is represented by the following formula (1). P(x)≒Σ i , i , i , i , i , i , i , , i , i , i c i exp{-(x - μ i )2 / σ i} ···(1) i is an integer from 1 to M. Σ i represents the symbol for calculating c i [[ID=2⃣0]]exp{-(x - μ i )^2 / σ i} when the value of i is each from 1 to M. c i is the coupling coefficient, μ i is the mean, and σ s i is the variance. In this case, for example, c i s , μ i s , and σ iThese are the fitting parameters (parameters calculated by curve fitting). In this embodiment, the case where M=5 (where the M phases to be estimated are hematite, magnetite, calcium ferrite, silicate slag, and vacancies) is illustrated. Therefore, Figure 3 shows an example of average μ1 to μ5. Average μ1 to μ5 corresponds to the pixel values ​​that show an upwardly convex peak in the frequency distribution 300. Note that the M phases to be estimated in the sintered ore are not limited to hematite, magnetite, calcium ferrite, silicate slag, and vacancies. Not all of these five phases are the M phases to be estimated in the sintered ore. For example, the M phases to be estimated in a sintered ore may include phases containing iron oxide (e.g., hematite, magnetite, and calcium ferrite). In this case, M=3, or M=4, with silicate slag and vacancies as other phases. Furthermore, the M phases to be estimated in a sintered ore may include phases other than hematite, magnetite, calcium ferrite, silicate slag, and vacancies.

[0034] When curve fitting the frequency distribution 300 with a Gaussian mixture distribution, the threshold derivation unit 102c uses, for example, M mean μ values. i Of these, two mutually adjacent mean μ j , μ j+1 Midpoint (calculated average value (=(μ) j +μ j+1 )÷2)) are used to determine the M-1 threshold Th for converting the image to be processed into an M-value image. j The following is derived: j is an integer between 1 and M-1. Figure 3 shows an example of thresholds Th1 to Th4 derived in this way. Alternatively, the threshold derivation unit 102 could use the mean μ of the Gaussian mixture distribution P(x). j from x j The integral value up to x of the function in question. j From the mean μ j+1 The integral value of and are equal to x j Th j It can also be derived as follows: Alternatively, instead of using a mixture of Gaussian distributions, the frequency distribution 300 may be curve-fitted using, for example, M independent Gaussian distributions. This method may be used, for example, when M peaks are clearly defined in the frequency distribution 300.

[0035] <<First multi-level image generation unit 102d>> The first multi-level image generation unit 102d generates M-1 threshold values ​​Th j Based on this, an image is generated as the first M-value image, which is an image in which the pixel values ​​of the image to be processed are converted into M values. The first M-value image is represented by M pixel values ​​(which are mutually distinct). For example, the pixel values ​​of the region in the image to be processed that have pixel values ​​less than or equal to the threshold Th1 are the first pixel values, and the threshold Th k Ultra, threshold Th k+1 The pixel value of the region having the following pixel values ​​is the k-th pixel value, and the threshold Th in the image to be processed is M-1 A first M-valued image is generated where the region beyond the specified value is the M-th pixel value. k is an integer between 1 and M-2 (inclusive). Note that if M is 2, the k-th pixel value does not exist.

[0036] The specific values ​​of the 1st to Mth pixel values ​​may be predetermined values. The relationship between pixel values ​​in the image to be processed and the relationship between pixel values ​​in the first M-level image may be the same or different. That is, even if the pixel values ​​of a region with a pixel value below threshold Th1 in the image to be processed are smaller than the pixel values ​​of other regions, it is not necessary to make the pixel value of that region (1st Mth pixel value) smaller than the pixel values ​​of the other regions (2nd to Mth pixel values) in the first M-level image; it may even be larger. Furthermore, the first M-level image may be a color image. In this case, the first multi-level image generation unit 102d may generate a color image represented by M types of colors, where the 1st to Mth pixel values ​​are the 1st to Mth colors, as the first M-level image.

[0037] Figure 4 illustrates an example of processing applied to an image to be processed. For the sake of clarity, all images in Figure 4 are shown in grayscale. Figure 4 shows that the first M-value image 420 is generated based on the image to be processed 410.

[0038] <<Second multi-level image generation unit 102e>> The second multi-level image generation unit 102e identifies regions that form the boundaries of multiple phases in the image 410 to be processed, and the pixel values ​​of the identified regions and M-1 threshold values ​​Th j Based on this, a second megaquantized image 430 is generated by correcting the pixel values ​​of the region that forms the boundary between multiple phases in the first megaquantized image 420. In the following explanation, the region that forms the boundary between multiple phases will be referred to as the phase boundary peripheral region, as needed. The phase boundary peripheral region is a region between multiple phases, where the spacing is narrower than the size assumed to be the size of the phases.

[0039] For example, if there is a large difference in brightness (luminance) between two phases separated by a phase boundary region, the brightness (luminance) in that phase boundary region may change in steps in the captured image. In this case, the brightness (luminance) in that phase boundary region may correspond to the brightness (luminance) of a different phase than the two phases. For example, while the luminance at the boundary between the hematite region and the vacancy region should normally correspond to the luminance of hematite or vacancy, this luminance may correspond to silicate slag, calcium ferrite, or magnetite.

[0040] Therefore, the second multi-level image generation unit 102e identifies the phase boundary region in the image 410 to be processed in order to change the pixel value of the phase boundary region to a pixel value corresponding to one of the phases among the multiple phases that sandwich the phase boundary region, if the pixel value of the phase boundary region is not a pixel value of one of the multiple phases that sandwich the phase boundary region (in the example above, if it is not a pixel value (luminance value) corresponding to hematite or void).

[0041] The process for identifying the region around the phase boundary in the image to be processed 410 may be, for example, the edge enhancement process described above. However, in this embodiment, an example is given in which adaptive binarization processing is included in the process for identifying the region around the phase boundary in the image to be processed 410, so that the effect of background brightness disturbance (non-uniformity) occurring throughout the entire field of view can be reduced (preferably eliminated) compared to the edge enhancement process. In the following description, the image to be processed after such processing will be referred to as the phase boundary identification binarized image, as necessary.

[0042] Figure 4 shows that a binarized image 440 for phase boundary identification is generated based on the image 410 to be processed. In Figure 4, the black areas in the binarized image 440 for phase boundary identification correspond to the area around the phase boundary. The second multi-level image generation unit 102e may change the pixel values ​​of blobs (blobs) of a preset size or smaller in the binarized image 440 for phase boundary identification to pixel values ​​corresponding to areas that are not the area around the phase boundary (in Figure 4, the black areas in the binarized image 440 for phase boundary identification may be changed to white areas).

[0043] A blob is a group of pixels that share the same pixel value. For example, the second multi-level image generation unit 102e defines a pixel with a pixel value corresponding to the phase boundary peripheral region and the surrounding pixels (8 pixels) (left-right, up-down, and diagonal directions) that also share the same pixel value for each pixel as part of the same blob. In this case, the second multi-level image generation unit 102e derives each blob as a separate phase boundary peripheral region.

[0044] In this embodiment, a case where the peripheral region of the phase boundary in the processing target image 410 is specified based on the binary image 440 for phase boundary specification is illustrated. For example, the second multi-valued image generation unit 102e specifies, as pixels belonging to the peripheral region of the phase boundary in the processing target image 410, the pixels corresponding to the pixels belonging to the peripheral region of the phase boundary in the binary image 440 for phase boundary specification among the pixels of the processing target image 410. Further, the second multi-valued image generation unit 102e performs labeling and assigns identification information (number) to each peripheral region of the phase boundary. In this embodiment, a case where the second multi-valued image generation unit 102e can perform processing for each peripheral region of the phase boundary in the processing target image 410 as described above is illustrated.

[0045] Then, the second multi-valued image generation unit 102e corrects the pixel values of the peripheral region of the phase boundary in the first M-valued image 420 based on the pixel values of the peripheral region of the phase boundary in the processing target image 410 and M - 1 threshold values Th j and.

[0046] For example, the second multi-valued image generation unit 102e determines, for each peripheral region of the phase boundary, to which of the M pixel value ranges defined based on the M - 1 threshold values Th j each pixel in the peripheral region of the phase boundary in the processing target image 410 belongs. In the following description, the pixel value range defined based on the M - 1 threshold values Th j is referred to as a multi-valued pixel value range as needed.

[0047] In the example shown in FIG. 3, the second multi-valued image generation unit 102e determines, for each peripheral region of the phase boundary, to which of the M multi-valued pixel value ranges r i (i is an integer from 1 to M) each pixel value of the peripheral region of the phase boundary in the processing target image 410 belongs. Specifically, the second multi-valued image generation unit 102e determines, for each peripheral region of the phase boundary, to which of the five multi-valued pixel value ranges r i (for example, r1≦Th1, Th1<r2≦Th2, Th2<r3≦Th3, Th3<r4≦Th4, Th4<r5) each pixel value of the peripheral region of the phase boundary in the processing target image 410 belongs.

[0048] Then, the second multi-level image generation unit 102e generates M multi-level pixel value ranges r to which each pixel value in the phase boundary peripheral region of the image 410 to be processed belongs. i Among them, the pixel value range r to which the largest number of pixel values ​​belong i The process of identifying the pixel values ​​of all pixels in the area surrounding the phase boundary is performed for each region around the phase boundary. The second multi-level image generation unit 102e determines that the pixel values ​​of all pixels in the area surrounding the phase boundary are within the multi-level pixel value range r identified in this way. i Assuming that the pixel values ​​are within the phase boundary, the pixel values ​​in the phase boundary peripheral region of the phase boundary identification binarized image 440 are collectively converted to M values ​​for each phase boundary peripheral region. Here, the M conversion is performed by converting the pixel values ​​of the image to be processed 410 into M multi-level pixel value ranges r as described above. i The range of pixel values ​​r used for multi-level conversion, selected from among them. i The process is carried out so that the pixel values ​​of the first M-valued image 420, obtained by M-valuing the image 410 to be processed, are within the same range (in the following explanation, such pixel values ​​will be referred to as representative first M-valued pixel values ​​as needed). For example, if the multi-valued pixel value range to which the most pixel values ​​belong in a certain phase boundary peripheral region is the multi-valued pixel value range r2, the second multi-valued image generation unit 102e changes the pixel values ​​of all pixels in the phase boundary peripheral region in the phase boundary identification binarized image 440 to the second pixel value in the first M-valued image 420. The second multi-valued image generation unit 102e also changes the pixel values ​​of pixels in regions of the phase boundary identification binarized image 440 other than the phase boundary peripheral region to a pixel value different from the i-th pixel value (where i is an integer from 1 to M). Note that, within the region of the phase boundary identification binarized image 440, the pixel values ​​of pixels in areas other than the region surrounding the phase boundary may be, for example, the pixel values ​​of the phase boundary identification binarized image 440. In the following explanation, the image generated by changing the pixel values ​​of the phase boundary identification binarized image 440 in the manner described above will be referred to as the phase boundary identification multilevel image, as necessary. Figure 4 shows that the phase boundary identification multilevel image 450 is generated based on the phase boundary identification binarized image 440.

[0049] The second multi-level image generation unit 102e then changes the pixel values ​​of the phase boundary peripheral regions in the first multi-level image 420 to the pixel values ​​of the phase boundary peripheral regions in the multi-level image 450 for phase boundary identification, for each phase boundary peripheral region. In this embodiment, we illustrate the case where the first multi-level image 420, whose pixel values ​​in the phase boundary peripheral regions have been changed in this way, is the second multi-level image. Figure 4 shows that the second multi-level image 430 is generated based on the first multi-level image 420 and the multi-level image 450 for phase boundary identification. The second multi-level image 430 has M pixel values. In the second multi-level image 430, M phases are represented by M mutually different pixel values ​​(1st to Mth pixel values). The 1st to Mth pixel values ​​in the first multi-level image 420 and the 1st to Mth pixel values ​​in the second multi-level image 430 are the same values. In the second M-quantized image 430, a group of regions with the same pixel value represents a single phase (individual phase). In the following explanation, a group of regions with the same pixel value in the M-quantized images (first M-quantized image 420 and second M-quantized image 430) will be referred to as an object as needed. If the second M-quantized image 430 is a color image, a group of regions of the same color represents a single phase (individual phase).

[0050] The method for generating the second megascale image is not limited to the method described above. For example, the image to be processed by converting the pixel values ​​in the area surrounding the phase boundary into M-values ​​all at once may be the image to be processed, for example, the image to be processed, rather than the binarized image 440 for phase boundary identification. Also, it is not necessary to convert the pixel values ​​in the area surrounding the phase boundary into M-values ​​all at once. In this case, for example, it may be determined whether each of the pixel values ​​in the area surrounding the phase boundary of the first M-value image 420 is a representative first M-value pixel value, and only the pixel values ​​in the area surrounding the phase boundary of the first M-value image 420 that are not representative first M-value pixel values ​​may be changed to representative first M-value pixel values.

[0051] Furthermore, for example, the second multi-level image generation unit 102e does not use the pixel values ​​of the phase boundary peripheral region in the image 410 to be processed as they are, but rather uses a value based on the pixel values ​​of the phase boundary peripheral region in the image 410 to be processed, and a multi-level pixel value range r iThe two may be compared. In this case, for example, the second multi-level image generation unit 102e may derive a representative value of the pixel values ​​in the phase boundary peripheral region of the image to be processed 410 for each phase boundary peripheral region. The representative value may be, for example, the mode, or a weighted average value (for example, a weighted average value with larger weight coefficients for pixels closer to the edge of the phase boundary peripheral region (pixels farther from the center)). In this case, the second multi-level image generation unit 102e may determine which multi-level pixel value range r the representative value of the pixel values ​​in the phase boundary peripheral region of the image to be processed 410 is suitable for. i It determines whether it belongs to the specified range. Then, the second multi-level image generation unit 102e determines that the pixel values ​​of all pixels in the region surrounding the phase boundary are within the determined multi-level pixel value range r. i Assuming that the pixel values ​​are within the phase boundary, the pixel values ​​in the phase boundary peripheral region of the phase boundary binarized image 440 are collectively converted to M values ​​for each phase boundary peripheral region, thereby generating a phase boundary multilevel image 450. Here, the M conversion is performed by converting the pixel values ​​of the image to be processed 410 to the multilevel pixel value range r to which the representative value of the pixel values ​​in the phase boundary peripheral region of the image to be processed 410 belongs. i The process is carried out so that the pixel values ​​of the first M-valued image 420, obtained by M-valuing the image 410 to be processed, are the same as those of the first M-valued image 420. The second multi-level image generation unit 102e then changes the pixel values ​​of the phase boundary peripheral regions in the first M-valued image 420 to the pixel values ​​of the same phase boundary peripheral regions in the phase boundary identification multi-level image 450, for each phase boundary peripheral region. For example, if a representative value of the pixel values ​​of a certain phase boundary peripheral region belongs to the multi-level pixel value range r2, the second multi-level image generation unit 102e changes the pixel values ​​of all pixels in that phase boundary peripheral region to the second pixel value in the first M-valued image 420. Even in this case, the image that is targeted for M-valuation of the pixel values ​​of the phase boundary peripheral regions all at once may be, for example, the image to be processed 410, rather than the phase boundary identification binarized image 440.

[0052] <<Feature extraction unit 102f>> The feature derivation unit 102f derives feature quantities for M phases based on the second M-valued image 430. The feature quantities represent, for example, features of at least one of the number, size, and shape of the M phases. In this embodiment, it is illustrated that the feature quantities for the M phases include at least one of the number rate, phase fraction rate, and aspect ratio. As mentioned above, in this embodiment, it is illustrated that the M phases to be estimated are hematite, magnetite, calcium ferrite, silicate slag, and vacancies. Therefore, feature quantities for hematite, magnetite, calcium ferrite, silicate slag, and vacancies are derived.

[0053] The feature derivation unit 102f may, for example, derive the ratio of the total number of objects representing a certain type of phase to the total number of objects included in the second M-valued image 430 as the count rate of that phase. Alternatively, the feature derivation unit 102f may derive the count rate by, for example, counting the number of grid point centers occupied by objects (phases) for each type of phase (M types of phases) using the point calculation method shown in JIS G0555:2020, and then deriving the number of M phases included in the second M-valued image 430 for each type of phase.

[0054] Furthermore, the feature derivation unit 102f may derive, for example, the ratio (area ratio) of the total area (total number of pixels) of objects representing a certain type of phase to the total area (total number of pixels) of objects included in the second M-valued image 430 as the phase fraction of that phase. Alternatively, the feature derivation unit 102f may derive the phase fraction by, for example, counting the number of grid point centers occupied by objects (phases) for each type of phase (M types of phases) using the point calculation method shown in JIS G0555:2020, and deriving the area of ​​each phase included in the second M-valued image 430 for each type of phase.

[0055] Furthermore, the feature derivation unit 102f may, for example, derive the bounding rectangle for each object included in the second M-valued image 430, and derive the aspect ratio of the bounding rectangle as the aspect ratio of the phase corresponding to the object. Alternatively, the feature derivation unit 102f may derive representative values ​​(e.g., arithmetic mean, median, mode) for each type of phase as the aspect ratio of the phase.

[0056] As described above, deriving the feature quantities of M phases is preferable because it provides more detailed information about the structure of each phase. However, the processing unit 100 does not necessarily need to derive the feature quantities of each phase. For example, detailed information about the structure of each phase can be obtained from the second M-valued image 430 alone. Furthermore, the aforementioned feature quantities can also be derived by the operator by referring to the second M-valued image 430. Therefore, the processing unit 100 does not need to include a feature quantity derivation unit 102f.

[0057] <<Output section 103>> The output unit 103 outputs at least one of the following: the second multi-level image 430 generated by the second multi-level image generation unit 102e, and information indicating the feature quantities of each phase derived by the feature quantity derivation unit 102f. In the following description, the information output by the output unit 103 will be referred to as output information.

[0058] In this embodiment, an example is given where the output unit 103 outputs output information to the output device 120. The output device 120 may include, for example, a computer display. In this case, the output device 120 displays the output information. The output device 120 may also include a communication device. In this case, the output device 120 transmits the output information to an external device. The output device 120 may also include a storage medium. In this case, for example, the output information stored in the storage medium of the output device 120 may be used by the processing unit 100 and the external device.

[0059] <Flowchart> Next, an example of a processing method performed using the processing unit 100 of this embodiment will be described with reference to the flowchart shown in Figure 5. The flowchart shown in Figure 5 is executed, for example, by the processor of the processing unit 100 reading and expanding a computer program stored in memory.

[0060] First, in step S501, the acquisition unit 101 acquires information including the captured image as input information. As mentioned above, in this embodiment, the acquisition unit 101 acquires an image as the captured image, which is an image of the polished surface (cross-section) of the sintered ore captured by the imaging means of an optical microscope.

[0061] Next, in step S502, the preprocessing unit 102a generates the image to be processed 410 by performing preprocessing on the captured image. As described above, in this embodiment, we illustrate a case in which the preprocessing unit 102a performs the following preprocessing steps: smoothing the captured image represented in grayscale, emphasizing the edges in the smoothed captured image, and combining the pixel values ​​of the captured image without edge emphasis with the pixel values ​​of the captured image with edge emphasis.

[0062] Next, in step S503, the frequency distribution derivation unit 102b derives the frequency distribution 300 of the pixel values ​​of the image 410 to be processed. Next, in step S504, the threshold derivation unit 102c determines M-1 thresholds Th for converting the image to be processed into an M-value image, based on the result of curve fitting the frequency distribution 300. j The following is derived. As mentioned above, in this embodiment, the threshold derivation unit 102c curve-fits the frequency distribution 300 with a Gaussian mixture distribution to obtain M mean μ i We derive the mean μ of two mutually adjacent values. j , μ j+1 The midpoint is the M-1 threshold Th required to convert the image to an M-value image. j An example of how to derive it is given.

[0063] Next, in step S505, the first multi-level image generation unit 102d generates M-1 threshold values ​​Th j Based on this, an image with M-values ​​of the pixel values ​​of the image to be processed 410 is generated as the first M-value image 420.

[0064] Next, in step S506, the second multi-level image generation unit 102e identifies the phase boundary peripheral region in the image to be processed 410, and the pixel values ​​of the identified phase boundary peripheral region and M-1 threshold values ​​Th j Based on this, an image is generated as the second M-value image 430 by correcting the pixel values ​​of the phase boundary peripheral region in the first M-value image 420. As described above, in this embodiment, the second multi-value image generation unit 102e generates a phase boundary identification binarized image 440 and a phase boundary identification multi-value image 450, and generates the second M-value image 430 by changing the pixel values ​​of the phase boundary peripheral region in the first M-value image 420 to the pixel values ​​of the phase boundary peripheral region in the phase boundary identification multi-value image 450 for each phase boundary peripheral region.

[0065] Next, in step S507, the feature derivation unit 102f derives M phase features based on the second M-valued image 430. As mentioned above, in this embodiment, we illustrate the case where the M phase features include at least one of the number rate, phase fraction rate, and aspect ratio. Finally, in step S508, the output unit 103 outputs output information. As described above, in this embodiment, we illustrate a case where the output information includes at least one of the second M-valued image 430 and information indicating the feature quantities of each phase.

[0066] <Summary> As described above, in this embodiment, the processing apparatus 100 performs image processing on a photographed image of a cross-section of a material composed of N phases bonded together. This processing includes deriving a frequency distribution of pixel values ​​of the image 410 to be processed, and determining M-1 threshold values ​​Th based on the result of curve fitting of the frequency distribution. j The process of deriving and M-1 thresholds Th jBased on this, a process is performed to generate an image as the first M-value image 420 in which the pixel values ​​of the image to be processed 410 have been converted to M values, and the pixel values ​​of the phase boundary peripheral region in the image to be processed 410 and M-1 thresholds Th j The process includes generating a second M-valued image 430 by correcting the pixel values ​​in the area surrounding the phase boundary in the first M-valued image 420 based on the above. Therefore, the composition of each phase of a substance composed of multiple combined phases can be easily and accurately estimated.

[0067] Furthermore, in this embodiment, the processing device 100 uses a grayscale image as the image to be processed 410. Therefore, information on the brightness (luminance) of each phase can be directly obtained from the pixel values. Thus, the processing to obtain the brightness (luminance) information of each phase can be performed in a short time (easily).

[0068] Furthermore, in this embodiment, the processing device 100 performs a process to smooth the captured image as a preprocessing step. Therefore, noise, including mosquito noise, contained in the captured image can be reduced. Thus, the composition of each phase of a substance composed of multiple combined phases can be estimated with higher accuracy.

[0069] Furthermore, in this embodiment, the processing device 100 performs preprocessing steps including a process to enhance the edges of the captured image, and a process to combine the pixel values ​​of the captured image without edge enhancement and the pixel values ​​of the captured image with edge enhancement. Therefore, the boundaries between multiple phases shown in the captured image can be made clearer. Thus, the composition of each phase of a substance composed of multiple combined phases can be estimated with higher accuracy.

[0070] Furthermore, in this embodiment, the processing unit 100 curve-fits the frequency distribution with a distribution function formed by combining M functions, each having one upward-convex peak, and then sets M-1 thresholds Th j This is derived. Therefore, even when the number of phase types to be estimated is large (when the value of M is large), the frequency distribution 300 can be curve-fitted quickly and with high accuracy.

[0071] Furthermore, in this embodiment, the processing device 100 generates a phase boundary identification binarized image 440 by performing adaptive binarization on the image to be processed 410, and identifies a region having pixel values ​​corresponding to the phase boundary peripheral region based on the phase boundary identification binarized image 440. Therefore, the phase boundary peripheral region can be identified with higher accuracy.

[0072] Furthermore, in this embodiment, the processing device 100 identifies one of the M multi-level pixel value ranges for multi-level conversion as the multi-level pixel value range corresponding to the phase boundary region, based on the pixel values ​​of the image 410 to be processed in the phase boundary region, and collectively changes the pixel values ​​of the phase boundary region in the first M-level converted image 420 to the pixel values ​​of the first M-level converted image 420 obtained by converting the image 410 to M-level values, assuming that the pixel values ​​of the image 410 to be processed are within the identified multi-level pixel value range.Therefore, it is possible to suppress the possibility that the pixel values ​​of the phase boundary region are not pixel values ​​corresponding to actual phases.As a result, the configuration of each phase of a substance composed of multiple combined phases can be estimated with higher accuracy.

[0073] Furthermore, in this embodiment, the processing device 100 derives M phase feature quantities based on the second M-valued image 430. Therefore, more detailed information about the composition of each phase of a substance composed of multiple combined phases can be easily obtained.

[0074] (Second Embodiment) A second embodiment will now be described. This embodiment adds further noise reduction processing to the processing of the image processing unit 102 in the first embodiment. Therefore, in the description of this embodiment, parts that are the same as in the first embodiment will be denoted by the same reference numerals as in Figures 1 to 5, and detailed explanations will be omitted.

[0075] <Processing device 600> Figure 6 shows an example of the functional configuration of the processing unit 600. In Figure 6, this embodiment illustrates a case where the processing unit 600 comprises an acquisition unit 101, an image processing unit 602, and an output unit 103. Some of the configurations of the image processing unit 602 differ from those of the processing unit 100 (image processing unit 102) of the first embodiment. An example of the configuration of the image processing unit 602 in this embodiment is described below.

[0076] In this embodiment, we illustrate a case where the image processing unit 602 includes a preprocessing unit 102a, a frequency distribution derivation unit 102b, a threshold derivation unit 102c, a first multilevel image generation unit 102d, a first void pixel value correction unit 602a, a feature quantity derivation unit 102f, a first phase pixel value correction unit 602b, a second multilevel image generation unit 602c, a second void pixel value correction unit 602d, and a second phase pixel value correction unit 602e. Note that the preprocessing unit 102a, frequency distribution derivation unit 102b, threshold derivation unit 102c, first multilevel image generation unit 102d, and feature quantity derivation unit 102f are the same as those described in the first embodiment.

[0077] <<First void pixel value correction unit 602a>> The first void pixel value correction unit 602a, when there is a region surrounded by a void in the first M-value image that is of a different phase than the void, changes the pixel value of the region of a different phase than the void included in the first M-value image to the pixel value corresponding to the void, based on the pixel value of the region of a different phase than the void in the image to be processed.

[0078] For example, the first void-internal pixel value correction unit 602a determines whether or not there is a region (object) surrounded by a void in the first multi-level image. For example, if the first multi-level pixel value correction unit 602a determines that an object not corresponding to a void is an object (phase) surrounded by a void if all the regions of an object not corresponding to a void are located inside the object corresponding to the void (inside the outer edge of the object corresponding to the void).

[0079] When a sintered ore is a material composed of N phases bonded together, the brightness in the captured image tends to be highest in the order of hematite, magnetite, calcium ferrite, silicate slag, and vacancies, as mentioned above. Therefore, it is sometimes preferable for the first vacancy pixel value correction unit 602a to consider relatively dark (low brightness) regions among objects surrounded by vacancies (i.e., objects within vacancies) as vacancies.

[0080] Therefore, in this embodiment, we illustrate a case in which the first void-internal pixel value correction unit 602a determines whether the representative value of the pixel value (luminance value) of the object surrounded by the void (i.e., the object inside the void) in the first M-level image is less than or equal to a threshold. The representative value may be, for example, the arithmetic mean, the median, the mode, the maximum value, the minimum value, or a weighted mean (for example, a weighted mean with larger weight coefficients for pixels closer to the center of the phase boundary peripheral region). In this embodiment, we illustrate a case in which the representative value is the arithmetic mean. The threshold may be, for example, the largest (brightest) value among the pixel values ​​corresponding to the silicate slug (the upper limit of the pixel value range r2 for multi-level conversion). In this case, the pixel value of the object corresponding to the silicate slug among the objects (phases) surrounded by the void is changed to the pixel value corresponding to the void. Alternatively, instead of determining whether the representative value of the pixel value of the object surrounded by the void is below a threshold, the first void-internal pixel value correction unit 602a may, for example, derive the number of pixels among the pixels of the object surrounded by the void whose pixel value is below a threshold, and then determine whether the derived number of pixels exceeds the threshold.

[0081] Then, the first void-internal pixel value correction unit 602a changes the pixel values ​​of objects surrounded by voids in the first M-value image whose representative pixel values ​​(luminance values) are below a threshold, to the pixel values ​​corresponding to the voids. In the following description, the first M-value image in which the pixel values ​​of objects surrounded by voids whose representative pixel values ​​(luminance values) are below a threshold have been changed in this manner will be referred to as the void-corrected first M-value image, as necessary.

[0082] Figure 7 illustrates an example of a process that changes the pixel values ​​of an object surrounded by a void. In the first megabyte-scaled image shown on the left side of Figure 7, the object 710 surrounded by the void is shown, and its pixel values ​​are changed to the pixel values ​​corresponding to the void. As a result, it no longer exists in the first megabyte-scaled image after correction of the void, shown on the right side of Figure 7 (at the end of the arrow).

[0083] <<First Phase Pixel Value Correction Unit 602b>> The first phase pixel value correction unit 602b, when it finds that in the first M-value image there is a region of a different phase surrounded by a single phase, changes the pixel value of the region of the different phase included in the first M-value image to the pixel value corresponding to the single phase, based on the size of the region of the different phase.

[0084] For example, the first phase pixel value correction unit 602b determines whether or not there is a region (object) surrounded by a single phase. For example, if the first phase pixel value correction unit 602b finds that all the regions of an object that does not correspond to a single phase are inside an object that corresponds to a single phase, it can determine that the object that does not correspond to a single phase is an object (phase) surrounded by that single phase.

[0085] In this embodiment, the first hole-corrected first M-value image may be generated by the first hole-corrected pixel value correction unit 602a. In this embodiment, when the first hole-corrected first M-value image is generated, the first phase-in-pixel value correction unit 602b determines whether or not there is another object surrounded by an object corresponding to a single phase in the first hole-corrected first M-value image. On the other hand, when the first hole-corrected first M-value image is not generated, the first phase-in-pixel value correction unit 602b determines whether or not there is another object surrounded by an object corresponding to a single phase in the first M-value image.

[0086] It may be preferable to consider an object corresponding to another smaller phase that exists (in isolation) within an object corresponding to a single phase as the object corresponding to that single phase. Therefore, in this embodiment, when a first M-valued image after correction within the void is generated, the first phase pixel value correction unit 602b determines whether the size of another (one) object surrounded by a single object in the first M-valued image after correction within the void is less than or equal to a threshold. On the other hand, when a first M-valued image after correction within the void is not generated, the first phase pixel value correction unit 602b determines whether the size of another (one) object surrounded by a single object in the first M-valued image is less than or equal to a threshold. The threshold is set in advance, for example, based on the size assumed to be the size as a single phase. The size of the object may be represented, for example, by the number of pixels having the pixel value corresponding to the object.

[0087] Then, the first phase in-phase pixel value correction unit 602b changes the pixel values ​​of the objects whose size is less than or equal to a threshold among the other objects surrounded by a single object in the first M-value image or the first M-value image after in-phase correction to the pixel values ​​corresponding to the single object. In the following description, the first M-value image in which the pixel values ​​of the other objects whose size is less than or equal to a threshold among the objects surrounded by a single object have been changed in this way will be referred to as the first M-value image after in-phase correction, as necessary.

[0088] <<Second multi-level image generation unit 602c, second void pixel value correction unit 602d, second phase pixel value correction unit 602e>> If the first M-valued image after correction within the void and the first M-valued image after correction within the phase are not generated, the second M-valued image generation unit 602c generates a second M-valued image by, for example, performing the processing described in the first embodiment. On the other hand, if the first M-valued image after correction within the void is generated, but the first M-valued image after correction within the phase is not generated, this embodiment illustrates a case where the second M-valued image generation unit 602c generates a second M-valued image using the first M-valued image after correction within the void. Furthermore, if the first M-valued image after correction within the phase is generated, this embodiment illustrates a case where the second M-valued image generation unit 602c generates a second M-valued image using the first M-valued image after correction within the phase (without using the first M-valued image after correction within the void).

[0089] The second void-internal pixel value correction unit 602d, when it detects that the second M-value image contains a region surrounded by a void that is of a different phase than the void, changes the pixel values ​​of the region of a different phase than the void in the second M-value image to the pixel values ​​corresponding to the void, based on the pixel values ​​of the region of a different phase than the void in the image to be processed. An example of the processing in the second void-internal pixel value correction unit 602d can be achieved by processing the first M-value image as the second M-value image in the processing of the first void-internal pixel value correction unit 602a. Therefore, a detailed explanation of the processing in the second void-internal pixel value correction unit 602d is omitted here. Note that the threshold used in the processing of the second void-internal pixel value correction unit 602d may be a different value from the threshold used in the processing of the first void-internal pixel value correction unit 602a. In the following explanation, the image generated by the second void-internal pixel value correction unit 602d will be referred to as the void-corrected second M-value image, as necessary.

[0090] The second phase pixel value correction unit 602e, in the second M-value image, if there is a region of a different phase surrounded by a single phase, changes the pixel value of the region of the different phase included in the second M-value image to the pixel value corresponding to the single phase, based on the size of the region of the different phase.

[0091] In this embodiment, the second hole-corrected pixel value correction unit 602d may generate a second M-value image. Therefore, in this embodiment, when a second M-value image after hole-correction is generated, the second phase-internal pixel value correction unit 602e determines whether or not there is another object surrounded by a single object in the second M-value image after hole-correction. On the other hand, when a second M-value image after hole-correction is not generated, the second phase-internal pixel value correction unit 602e determines whether or not there is another object surrounded by a single object in the second M-value image.

[0092] An example of the processing in the second-phase intra-pixel value correction unit 602e can be achieved by processing the first-phase intra-pixel value correction unit 602b to convert the first M-value image and the first M-value image after intra-void correction to the second M-value image and the second M-value image after intra-void correction, respectively. Therefore, a detailed explanation of the processing in the second-phase intra-pixel value correction unit 602e is omitted here. Note that the threshold used in the processing of the second-phase intra-pixel value correction unit 602e may be a different value from the threshold used in the processing of the first-phase intra-pixel value correction unit 602b. In the following explanation, the image generated by the second-phase intra-pixel value correction unit 602e will be referred to as the second M-value image after intra-phase correction, as necessary.

[0093] <<Feature Derivation Unit 602f>> If the second M-valued image after correction within the void and the second M-valued image after correction within the phase are not generated, the feature quantity derivation unit 602f derives M phase feature quantities based on the second M-valued image by performing the processing described in the first embodiment. On the other hand, if the second M-valued image after correction within the void is generated, but the second M-valued image after correction within the phase is not generated, this embodiment exemplifies a case where the feature quantity derivation unit 602f derives M phase feature quantities using the second M-valued image after correction within the void. Furthermore, if the second M-valued image after correction within the phase is generated, this embodiment exemplifies a case where the feature quantity derivation unit 602f derives M phase feature quantities using the second M-valued image after correction within the phase (without using the second M-valued image after correction within the void).

[0094] <Flowchart> Next, an example of a processing method performed using the processing unit 600 of this embodiment will be described with reference to the flowchart shown in Figure 8. The flowchart shown in Figure 8 is executed, for example, by the processor of the processing unit 600 reading and expanding a computer program stored in memory.

[0095] An example of the processing in steps S801 to S805 in Figure 8 is the same as the processing in steps S501 to S505 in Figure 5. When the first M-valued image is generated in step S805, the processing in step S806 is performed.

[0096] In step S806, the first void-in-pixel value correction unit 602a determines whether or not there are objects surrounded by voids in the first M-value image. If, as a result of this determination, there are no objects surrounded by voids in the first M-value image (NO in step S806), the processes in steps S807 to S808 are not performed, and the process in step S809, described later, is performed. On the other hand, if there is at least one object surrounded by voids in the first M-value image (YES in step S806), the process in step S807 is performed.

[0097] In step S807, the first void pixel value correction unit 602a determines whether or not, in the first M-value image, there are objects surrounded by voids whose representative pixel value (luminance value) is below a threshold.

[0098] As a result of this determination, if in the first M-value image there are no objects surrounded by voids whose representative pixel value (luminance value) is below the threshold (the answer in step S807 is NO), the process in step S808 is not performed, and the process in step S809, described later, is performed. On the other hand, if in the first M-value image there is at least one object surrounded by voids whose representative pixel value (luminance value) is below the threshold (the answer in step S807 is YES), the process in step S808 is performed.

[0099] In step S808, the first void-internal pixel value correction unit 602a generates a first void-internal corrected void-internal image in which, in the first void-internal image, the pixel values ​​of objects surrounded by voids whose representative pixel values ​​(luminance values) are below a threshold are changed to the pixel values ​​corresponding to the voids.

[0100] Next, in step S809, the first phase pixel value correction unit 602b determines whether, if the processing in step S808 is performed and the first M-value image after correction within the void is generated, there is another object surrounded by an object corresponding to a single phase (an area of ​​another phase surrounded by a single phase) in the first M-value image after correction within the void. On the other hand, if the processing in step S808 is not performed, the first phase pixel value correction unit 602b determines whether, in the first M-value image, there is another object surrounded by an object corresponding to a single phase (an area of ​​another phase surrounded by a single phase).

[0101] If, as a result of this determination, there are no other objects surrounded by the object corresponding to the single phase (the result is NO in step S809), then the processing in steps S810 to S811 is not performed, and the processing in step S812, described later, is performed. On the other hand, if there is at least one other object surrounded by the object corresponding to the single phase (the result is NO in step S809), then the processing in step S810 is performed.

[0102] In step S810, if the first phase pixel value correction unit 602b performs the processing in step S808 and generates the first M-value image after correction within the void, it determines whether or not there is an object in the first M-value image after correction within the void that is surrounded by a single object and has a size less than or equal to a threshold. On the other hand, if the processing in step S808 is not performed, the first phase pixel value correction unit 602b determines whether or not there is an object in the first M-value image that is surrounded by an object corresponding to a single phase and has a size less than or equal to a threshold.

[0103] As a result of this determination, if there are no other objects surrounded by the objects corresponding to a single phase that are less than or equal to the threshold in size (the result is NO in step S810), the process in step S811 is not performed and the process in step S812, described later, is performed. On the other hand, if there is at least one other object surrounded by the single object that is less than or equal to the threshold in size (the result is YES in step S810), the process in step S811 is performed.

[0104] In step S811, if the first phase-internal pixel value correction unit 602b has performed the processing in step S808 and generated the first M-value image after inter-void correction, it has changed the pixel values ​​of objects in the first M-value image after inter-void correction that are surrounded by objects corresponding to a single phase and whose size is less than or equal to a threshold, to the pixel values ​​corresponding to the single phase, and has generated an image as the first M-value image after inter-phase correction. On the other hand, if the processing in step S808 has not been performed, the first phase-internal pixel value correction unit 602b has changed the pixel values ​​of objects in the first M-value image that are surrounded by objects corresponding to a single phase and whose size is less than or equal to a threshold, to the pixel values ​​corresponding to the single phase, and has generated an image as the first M-value image after inter-phase correction.

[0105] Next, in step S812, if the processing in steps S808 and S811 was not performed, the second multi-level image generation unit 602c identifies the phase boundary peripheral region in the image to be processed, and the pixel values ​​of the identified phase boundary peripheral region and M-1 threshold values ​​Th j Based on this, an image in which the pixel values ​​of the phase boundary peripheral region in the first M-level image have been corrected is generated as the second M-level image. On the other hand, if the processing in step S808 is performed but the processing in step S811 is not performed, and the first M-level image after correction within the void is generated, but the first M-level image after correction within the phase is not generated, the second multilevel image generation unit 602c identifies the phase boundary peripheral region in the first M-level image after correction within the void, and uses the pixel values ​​of the identified phase boundary peripheral region and M-1 threshold values ​​Th jBased on this, an image is generated as a second M-level image in which the pixel values ​​of the phase boundary peripheral region in the first M-level image after correction within the void are corrected. Also, when the processing in step S811 is performed and the first M-level image after correction within the phase is generated, the second multilevel image generation unit 602c identifies the phase boundary peripheral region in the first M-level image after correction within the phase, and uses the pixel values ​​of the identified phase boundary peripheral region and M-1 thresholds Th j Based on this, a second M-valued image is generated by correcting the pixel values ​​in the area surrounding the phase boundary in the first M-valued image after correction within the phase hole.

[0106] Next, in step S813, the second void pixel value correction unit 602d determines whether or not there are objects surrounded by voids in the second M-value image. If, as a result of this determination, there are no objects surrounded by voids in the second M-value image (NO in step S813), the processing in steps S814 to S815 is not performed, and the processing in step S816, described later, is performed. On the other hand, if there is at least one object surrounded by voids in the second M-value image (YES in step S813), the processing in step S814 is performed.

[0107] In step S814, the second void pixel value correction unit 602d determines whether or not, in the second M-value image, there are objects surrounded by voids whose representative pixel value (luminance value) is below a threshold.

[0108] As a result of this determination, if in the second M-value image there are no objects surrounded by voids whose representative pixel value (luminance value) is below the threshold (the answer in step S814 is NO), the process in step S815 is not performed, and the process in step S816, described later, is performed. On the other hand, if in the second M-value image there is at least one object surrounded by voids whose representative pixel value (luminance value) is below the threshold (the answer in step S814 is YES), the process in step S815 is performed.

[0109] In step S815, the second void-internal pixel value correction unit 602d generates a second void-internal corrected void-internal image in which, in the second void-internal image, the pixel values ​​of objects surrounded by voids whose representative pixel values ​​(luminance values) are below a threshold are changed to the pixel values ​​corresponding to the voids.

[0110] Next, in step S816, the second phase pixel value correction unit 602e determines whether, if the processing in step S815 is performed and the second M-value image after correction within the void is generated, there is another object surrounded by an object corresponding to a single phase (an area of ​​another phase surrounded by a single phase) in the second M-value image after correction within the void. On the other hand, if the processing in step S815 is not performed, the second phase pixel value correction unit 602e determines whether, in the second M-value image, there is another object surrounded by an object corresponding to a single phase (an area of ​​another phase surrounded by a single phase).

[0111] If, as a result of this determination, there are no other objects surrounded by the object corresponding to the single phase (the result is NO in step S816), then the processing in steps S817 to S818 is not performed, and the processing in step S819, described later, is performed. On the other hand, if there is at least one other object surrounded by the object corresponding to the single phase (the result is NO in step S816), then the processing in step S817 is performed.

[0112] In step S817, if the processing in step S815 is performed and a second M-value image after correction within the void is generated, the second M-value image after correction within the void contains an object whose size is less than or equal to a threshold, as another object surrounded by a single object. On the other hand, if the processing in step S815 is not performed, the second M-value correction unit 602e determines whether or not there is an object whose size is less than or equal to a threshold, as another object surrounded by an object corresponding to a single phase, in the second M-value image.

[0113] As a result of this determination, if there are no other objects surrounded by the object corresponding to the single phase that are less than or equal to the threshold in size (the answer is NO in step S817), the process in step S818 is not performed, and the process in step S812, described later, is performed. On the other hand, if there is at least one other object surrounded by the single object that is less than or equal to the threshold in size (the answer is YES in step S817), the process in step S818 is performed.

[0114] In step S818, if the processing in step S815 is performed and a second M-value image after correction within the void is generated, the second M-value image after correction within the void is generated by changing the pixel values ​​of objects whose size is below a threshold among other objects surrounded by objects corresponding to a single phase to the pixel values ​​corresponding to that single phase. On the other hand, if the processing in step S815 is not performed, the second M-value image after correction within the void is generated by changing the pixel values ​​of objects whose size is below a threshold among other objects surrounded by objects corresponding to a single phase to the pixel values ​​corresponding to that single phase.

[0115] Next, in step S819, if the processes in steps S815 and S818 were not performed, the feature derivation unit 602f derives M phase features based on the second M-valued image 430. On the other hand, if the process in step S815 was performed but the process in step S818 was not performed, and the second M-valued image after correction within the void was generated, but the second M-valued image after correction within the phase was not generated, the feature derivation unit 602f derives M phase features using the second M-valued image after correction within the void. Also, if the process in step S818 was performed and the second M-valued image after correction within the phase was generated, the feature derivation unit 602f derives M phase features using the second M-valued image after correction within the phase.

[0116] Next, in step S820, the output unit 103 outputs output information. In this embodiment, as in the first embodiment, we will illustrate a case where the output information includes at least one of the second M-valued image 430 and information indicating the feature quantities of each phase. Note that the processing in step S820 is the same as the processing in step S808.

[0117] <Summary> As described above, in this embodiment, the processing device 600 generates a first M-value image after correction within voids, a first M-value image after correction within phases, a second M-value image after correction within voids, and a second M-value image after correction within phases. Therefore, within voids, phases (objects) having pixel values ​​close to the pixel values ​​corresponding to the voids can be considered noise and removed. Also, small phases (objects) existing in isolation within a certain phase can be considered noise and removed. Thus, the composition of each phase of a substance composed of multiple combined phases can be estimated with higher accuracy.

[0118] Furthermore, it is preferable to generate all of the following: the first M-value image after void correction, the first M-value image after phase correction, the second M-value image after void correction, and the second M-value image after phase correction, as this allows for a more reliable reduction of noise contained in the image 410 to be processed. However, at least one of the first M-value image after void correction, the first M-value image after phase correction, the second M-value image after void correction, and the second M-value image after phase correction does not need to be generated. For example, if a substance composed of N phases bonded together is a substance in which the presence of voids is not expected, the first M-value image after void correction and the second M-value image after void correction do not need to be generated.

[0119] (Third embodiment) Next, a third embodiment will be described. In the first embodiment, based on the curve fitting result of the frequency distribution 300, M-1 thresholds Th are used to convert the image to be processed into an M-value image. j We derive the following. In this case, there are M-1 thresholds Th j If it is not possible to derive the following, or if there are M-1 thresholds Th jIt is possible that at least one of the thresholds may not be of an appropriate value. Therefore, in this embodiment, M-1 thresholds Th j This example illustrates a case where at least one of the thresholds can be set to a threshold specified by the operator. Thus, this embodiment provides M-1 thresholds Th compared to the first embodiment. j This embodiment adds configurations and processes that modify the original. Therefore, in the description of this embodiment, parts that are the same as those in the first embodiment will be denoted by the same reference numerals as in Figures 1 to 5, and detailed explanations will be omitted.

[0120] <Processing device 900> Figure 9 shows an example of the functional configuration of the processing unit 900. In Figure 6, this embodiment illustrates a case where the processing unit 900 comprises an acquisition unit 101, an image processing unit 902, and an output unit 103. Some of the configurations of the image processing unit 902 differ from those of the processing unit 100 (image processing unit 102) of the first embodiment. An example of the configuration of the image processing unit 902 in this embodiment is described below.

[0121] In this embodiment, we illustrate a case where the image processing unit 602 includes a preprocessing unit 102a, a frequency distribution derivation unit 102b, a second multilevel image generation unit 102e, a feature quantity derivation unit 102f, a threshold derivation unit 902a, and a first multilevel image generation unit 902b. The preprocessing unit 102a, the frequency distribution derivation unit 102b, the second multilevel image generation unit 102e, and the feature quantity derivation unit 102f are the same as those described in the first embodiment.

[0122] <<Threshold Derivation Unit 902a>> The threshold derivation unit 902a, similar to the threshold derivation unit 102c in the first embodiment, calculates a threshold Th for converting the image to be processed into an M-value image based on the result of curve fitting of the frequency distribution 300. j The mean μ is derived. In this embodiment as in the first embodiment, the threshold derivation unit 902a curve-fits the frequency distribution 300 with a Gaussian mixture distribution P(x) to obtain the mean μ. i We derive the mean μ of two mutually adjacent values. j , μ j+1The midpoint is set to the threshold Th for converting the image to be processed into an M-value image. j An example of how to derive the threshold value is shown. The processing of the threshold value derivation unit 902a up to this point is the same as the processing of the threshold value derivation unit 102c in the first embodiment.

[0123] The threshold derivation unit 902a of this embodiment calculates M-1 thresholds Th if the result of curve fitting of the frequency distribution 300 does not satisfy predetermined conditions. j At least one of these thresholds is set to the threshold specified by the operator. In the following description, the given conditions will be referred to as multi-level conditions as needed.

[0124] To achieve this, for example, if the curve fitting of the frequency distribution 300 does not satisfy the conditions for multi-level recognition, the output device 120 may display a screen including a GUI (Graphical User Interface) for accepting the modified threshold. In the following description, this screen will be referred to as the threshold change screen as needed.

[0125] In this case, the threshold derivation unit 902a determines whether the multi-leveling feasibility condition is met based on the result of curve fitting of the frequency distribution 300. The multi-leveling feasibility derivation condition is the threshold Th that allows the image to be processed 410 to be multi-leveled (M-leveled). j The condition is that the multi-leveling is possible, for example, by a threshold Th derived by curve fitting the frequency distribution 300. j The condition may include that the number of elements is M-1. Furthermore, the multi-leveling condition is derived from a threshold Th obtained by curve fitting the frequency distribution 300. j When these are arranged in ascending order of value, the two thresholds Th are adjacent to each other in order. j (Two consecutive thresholds Th for which the value of j is j The condition may include that the absolute value of the difference between ) is greater than or equal to a predetermined value. Furthermore, the condition for multi-level classification is derived from a threshold Th obtained by curve fitting the frequency distribution 300. jThe condition may include that the maximum value of is within a predetermined range or less than or equal to a predetermined value. Furthermore, the multi-leveling condition may include a threshold Th derived by curve fitting of the frequency distribution 300. j The condition may include that the minimum value of is within a predetermined range or greater than or equal to a predetermined value. The conditions for multi-level conversion are pre-set in the processing unit 900.

[0126] Then, if the threshold derivation unit 902a does not meet the conditions for multi-level conversion, it instructs the output unit 103 to display the threshold change screen. Based on this instruction, the output unit 103 displays the threshold change screen on the computer display provided by the output device 120. Thus, in this embodiment, the threshold derivation unit 902a instructs the output unit 103 to display the threshold change screen when the result of curve fitting the frequency distribution 300 does not satisfy a predetermined condition (multilevel derivation condition). However, this is not always necessary. That is, the threshold derivation unit 902a may instruct the output unit 103 to display the threshold change screen regardless of whether the result of curve fitting the frequency distribution 300 satisfies a predetermined condition or not. In this case, the operator can use the threshold Th derived by curve fitting the frequency distribution 300. j This can be checked each time.

[0127] Figure 10 shows an example of the threshold change screen 1000. Figure 10 illustrates a case where the threshold change screen 1000 includes a threshold derivation result display unit 1010 and a threshold reception unit 1020. For the sake of explanation, it is assumed here that the frequency distribution 300 shown in Figure 3 is a frequency distribution that does not satisfy the conditions for multi-level recognition.

[0128] The threshold derivation result display unit 1010 displays the threshold Th derived by curve fitting the frequency distribution 300. jThe information to be shown is displayed. In FIG. 10, in the threshold value derivation result display unit 1010, the average μ i (average of the mixture Gaussian distribution) and the threshold value Th j are shown superimposed as an example of the case where the information is displayed. Also in this embodiment, as in the first embodiment, the case where M = 5 (when the M phases to be estimated are hematite, magnetite, calcium ferrite, silicate slag, and pores) is exemplified. Therefore, in order to generate the first M-valued image, 4 (= M - 1) threshold values Th j are required. In FIG. 10, an example is shown where four threshold values Th j are derived by curve fitting the frequency distribution 300.

[0129] On the other hand, when M - 1 threshold values Th j are not derived by curve fitting the frequency distribution 300, for example, the non-derived threshold value Th j may be displayed on the threshold value change screen 1000 (threshold value derivation result display unit 1010, threshold value reception unit 1020) as a preset value (for example, the minimum value (0.00 in the example shown in FIG. 10)).

[0130] Note that in FIG. 10, the average μ i derived by curve fitting the frequency distribution 300 is shown by a dashed line. As illustrated in FIG. 10, by superimposing the threshold value Th j derived by curve fitting the frequency distribution 300 on the frequency distribution 300 (graph), it is possible to make it easier for the operator to determine how to change the threshold value Th j .

[0131] In the threshold value reception unit 1020, a GUI for receiving the change value of at least one threshold value Th j among the M - 1 threshold values Th j is displayed. In FIG. 10, the M - 1 threshold values Thj An example is given of the case where a GUI for receiving a change value of the threshold value is displayed on the threshold value receiving unit 1020. However, it is not necessarily required to be like this. For example, a change value of some of the M-1 threshold values Th j (for example, only the threshold values not derived by the threshold value derivation unit 102c) may be received. Also, in FIG. 10, by changing the positions of the sliders 1021a to 1021d, the threshold values Th j (Th1 to Th4) shown on the left side of the sliders are changed. Also, in FIG. 10, an example is given of the case where the values of the threshold values Th j corresponding to the positions of the sliders 1021a to 1021d are displayed in the threshold value display columns 1022a to 1022d. Also, an example is given of the case where the positions of the threshold values Th j displayed in the threshold value derivation result display unit 1010 are changed according to the positions of the sliders 1021a to 1021d. Also, in FIG. 10, an example is given of the case where the OK buttons 1023a to 1023d to be pressed when determining the change values of the M threshold values Th j are displayed. Note that the GUI for changing the value of the threshold value Th j is not limited to a slider, and for example, a text box may also be used.

[0132] The operator operates the sliders 1021a to 1021d by using the input device 110. The acquisition unit 101 acquires information indicating the content of the operation and outputs it to the threshold value derivation unit 902a. Thus, in the present embodiment, an example is given of the case where the input information includes information indicating the content of the operation on the threshold value change screen 1000 (sliders 1021a to 1021d). The threshold value derivation unit 902a specifies the change value of the threshold value Th j based on the information, and based on the specified change value, instructs the output unit 103 to update the threshold value change screen 1000. In the threshold value change screen 1000, the information to be updated is the threshold value Th in the threshold value derivation result display unit 1010 jThese are the position of the sliders 1021a to 1021d and the values ​​displayed in the threshold display fields 1022a to 1022d. The output unit 103 displays the updated threshold change screen 1000 on the output device 120 (computer display). Thus, in this embodiment, an example is given in which the output information includes information for displaying the threshold change screen 1000.

[0133] The operator uses the input device 110 to operate (press) the OK buttons 1023a to 1023d, thereby determining the threshold Th j The acquisition unit 101 acquires information indicating the content of the operation and outputs it to the threshold derivation unit 902a. Based on this information, the threshold derivation unit 902a determines the threshold Th corresponding to the position of the sliders 1021a to 1021d at the time the OK buttons 1023a to 1023d were pressed. j Identify the value of the identified threshold Th j The value of is the threshold Th for generating the M-value image. j The value is determined as follows. Note that if the OK buttons 1023a to 1023d are pressed without the sliders 1021a to 1021e being operated, the threshold Th corresponding to the unoperated slider is determined. j This will not be changed.

[0134] Then, the threshold derivation unit 902a uses the threshold Th determined in the manner described above. j This is output to the first multi-level image generation unit 902b. On the other hand, if the result of curve fitting the frequency distribution 300 satisfies the conditions for multi-level recognition, the threshold change screen 1000 is not displayed in this embodiment. In this case, the threshold derivation unit 902a calculates the threshold Th derived by curve fitting the frequency distribution 300. j This is output to the first multi-level image generation unit 902b.

[0135] <<First Multi-level Image Generation Unit 902b>> The first multi-level image generation unit 902b generates M-1 threshold values ​​Th jBased on this, an image in which the pixel values ​​of the image to be processed have been converted to M values ​​is generated as the first M-value image. The first multi-level image generation unit 902b of this embodiment converts the value of the threshold Th specified by the operator via the threshold change screen 1000 to the threshold Th j Regarding the specified threshold Th, j This is used. On the other hand, the threshold Th that was not changed by the threshold derivation unit 902a is used. j The threshold Th was derived by curve fitting the frequency distribution of 300. j Use this. Other than this (the threshold Th to use) j The processing of the first multi-level image generation unit 902b (except for the difference in the first multi-level image generation unit 902b) is the same as the processing of the first multi-level image generation unit 102d in the first embodiment.

[0136] <Flowchart> Next, an example of a processing method performed using the processing unit 900 of this embodiment will be described with reference to the flowchart shown in Figure 11. The flowchart shown in Figure 11 is executed, for example, by the processor of the processing unit 900 reading and expanding a computer program stored in memory.

[0137] An example of the processing in steps S1101 to S1104 in Figure 11 is the same as the processing in steps S501 to S504 in Figure 5. When the first M-valued image is generated in step S1104, the processing in step S1105 is performed.

[0138] In step S1104, the threshold Th jOnce the value is derived, in step S1105, the threshold derivation unit 902a determines whether the result of curve fitting the frequency distribution 300 satisfies the multi-level classification condition. If the result of this determination satisfies the multi-level classification condition (YES in step S1105), the processes in steps S1106 to S1108 are not performed, and the process in step S1109, described later, is performed. Note that if the threshold change screen 1000 is to be displayed regardless of whether the result of curve fitting the frequency distribution 300 satisfies the multi-level classification condition, the process in step S1105 is not performed, and the process in step S1106 is performed after the completion of the process in step S1104.

[0139] On the other hand, if the result of curve fitting the frequency distribution 300 does not satisfy the conditions for multi-level classification (the result is NO in step S1105), the processes in steps S1106 to S1108 are performed. In step S1106, the threshold derivation unit 902a instructs the output unit 103 to display the threshold change screen 1000. Based on this instruction, the output unit 103 displays the threshold change screen 1000 on the computer display provided by the output device 120.

[0140] Next, in step S1107, the threshold derivation unit 902a determines a threshold Th for generating an M-value image based on the content of the operator's operation on the threshold change screen 1000. j The unit waits until it can determine the value of the threshold. In the example shown in Figure 10, the threshold derivation unit 902a waits until the OK buttons 1023a to 1023d are pressed. Then, the threshold Th for generating the M-value image is determined. j When it becomes possible to determine the value (if YES in step S1107), the process in step S1108 is performed.

[0141] In step S1108, the threshold derivation unit 902a determines the threshold Th for generating the M-value image based on the content of the operator's operation on the threshold change screen 1000. jThe value is determined. In the example shown in Figure 10, the threshold derivation unit 902a determines the threshold Th corresponding to the position of sliders 1021a to 1021d when the OK buttons 1023a to 1023d are pressed. j The value of is the threshold Th for generating the M-value image. j The value is determined. Then, the process in step S1109 is performed.

[0142] In step S1109, the first multi-level image generation unit 902b generates M-1 threshold values ​​Th j Based on this, an image in which the pixel values ​​of the image to be processed have been converted to M values ​​is generated as the first M-value image 420. If NO is determined in step S1104, the same as in the first embodiment, M-1 threshold values ​​Th derived in step S1104 are generated. j (M-1 thresholds Th derived by curve fitting of frequency distribution 300) j ) is used. On the other hand, if YES is determined in step S1104, M-1 thresholds Th are determined in step S1108 based on the content of the operator's operation on the threshold change screen 1000. j This is used. Other than this (the threshold used Th j The processing in step S1109 (except that it differs from the other step) is the same as the processing in step S505 of the first embodiment. The subsequent steps S1110 to S1112 are the same as the steps S506 to S508 in Figure 5.

[0143] <Summary> As described above, in this embodiment, the processing device 900 has M-1 threshold values ​​Th j At least one of the thresholds is set to the threshold specified by the operator. Thus, there are M-1 thresholds Th j This can be set to the threshold intended by the operator. In addition, in this embodiment, if the result of curve fitting the frequency distribution 300 does not satisfy the conditions for multi-level classification, the processing device 900 sets M-1 thresholds Th jAt least one of the thresholds is set to the threshold specified by the operator. Thus, by curve fitting the frequency distribution 300, M-1 thresholds Th are obtained. j If the appropriate threshold Th is not derived, j Even if this is not derived, the threshold Th corresponding to the brightness (pixel value) of the M phases displayed in the image 410 to be processed is used. j This allows for setting the configuration of each phase of a substance composed of multiple combined phases, thereby enabling more accurate estimation of the composition of each phase. The method of this embodiment may also be applied to a second embodiment.

[0144] (Other embodiments) The embodiments of this disclosure described above can be implemented by a computer executing a program. Furthermore, a computer-readable recording medium on which the program is stored, and a computer program product such as the program itself, can also be applied as embodiments of this disclosure. Examples of recording media include flexible disks, hard disks, optical disks, magneto-optical disks, CD-ROMs, magnetic tapes, non-volatile memory cards, ROMs, etc. Moreover, the embodiments of this disclosure may be implemented by a PLC (Programmable Logic Controller) or by dedicated hardware such as an ASIC (Application Specific Integrated Circuit). Furthermore, the embodiments of this disclosure described above are merely examples of concrete implementations of this disclosure, and the technical scope of this disclosure should not be interpreted as being limited by them. In other words, this disclosure can be implemented in various ways without departing from its technical concept or its main features.

[0145] Furthermore, the disclosure of the above embodiments is as follows, for example. [Disclosure 1] An acquisition unit that acquires an image of a cross-section of a material composed of N phases bonded together, using a microscope equipped with an imaging means. An image processing unit that performs image processing on the captured image, Equipped with, The aforementioned image processing unit, A frequency distribution derivation unit that derives the frequency distribution of pixel values ​​in the image to be processed, A threshold derivation unit that derives M-1 thresholds for converting the image to be processed into an M-value image based on the results of curve fitting the frequency distribution, A first multi-level image generation unit generates an image as a first M-level image in which the pixel values ​​of the image to be processed are converted to M-level values ​​based on the M-1 thresholds, A second multi-level image generation unit identifies regions that form the boundaries of multiple phases in the image to be processed, and generates a second multi-level image by correcting the pixel values ​​of the regions that form the boundaries of multiple phases in the first multi-level image based on the pixel values ​​of the identified regions and the M-1 thresholds. It has, N is an integer greater than or equal to 2, M is an integer between 2 and N, The processing apparatus wherein the image to be processed is the captured image, or an image that has been preprocessed from the captured image. [Disclosure 2] The image processing apparatus according to Disclosure 1, wherein the image to be processed is an image represented in grayscale. [Disclosure 3] The image processing unit includes a preprocessing unit that performs preprocessing on the captured image, The apparatus according to disclosure 1 or 2, wherein the preprocessing includes a process for smoothing the captured image. [Disclosure 4] The apparatus according to disclosure 1 or 2, wherein the preprocessing includes a process for enhancing the edges of the captured image, and a process for combining the pixel values ​​of the captured image without edge enhancement and the pixel values ​​of the captured image with edge enhancement. [Disclosure 5] The processing apparatus according to disclosure 4, wherein the synthesis includes calculating the average value of the pixel values ​​of the captured image without edge enhancement and the pixel values ​​of the captured image with edge enhancement. [Disclosure 6] The processing apparatus according to any one of disclosures 1 to 5, wherein the threshold derivation unit derives M-1 thresholds based on the result of curve fitting the frequency distribution with a distribution function obtained by combining M functions, each having one upward-convex peak. [Disclosure 7] The processing apparatus according to any one of disclosures 1 to 6, wherein the second multi-level image generation unit generates a binarized image by performing adaptive binarization on the image to be processed, and identifies regions that are boundaries between multiple phases in the image to be processed based on the binarized image. [Disclosure 8] The processing apparatus according to any one of disclosures 1 to 7, comprising: identifying one of M pixel value ranges determined based on M-1 thresholds as the pixel value range corresponding to the region of the processing target image based on the pixel values ​​of the region of the processing target image in the region that forms the boundary of the plurality of phases; and collectively changing the pixel values ​​of the region in the first M-value image to the pixel values ​​of the first M-value image obtained by M-value conversion of the processing target image assuming that the pixel values ​​of the processing target image are within the identified pixel value range. [Disclosure 9] The processing apparatus according to any one of disclosures 1 to 8, wherein the threshold derivation unit sets at least one of the M-1 thresholds to a threshold specified by the operator. [Disclosure 10] The apparatus according to disclosure 9, wherein the threshold derivation unit sets at least one of the M-1 thresholds to a threshold specified by the operator if the result of curve fitting the frequency distribution does not satisfy predetermined conditions. [Disclosure 11] The N phases include gas present in the vacancies of the substance. The processing apparatus according to any one of disclosures 1 to 10, wherein the image processing apparatus has a first void pixel value correction unit that, in the first M-value image, if there is a region of a different phase from the gas present in the void as a region surrounded by a region corresponding to the void, changes the pixel value of the region of the different phase from the gas included in the first M-value image to the pixel value corresponding to the gas, based on the pixel value of the region of the different phase from the gas in the image to be processed. [Disclosure 12] The N phases include gas present in the vacancies of the substance. The processing apparatus according to any one of disclosures 1 to 11, wherein the image processing apparatus has a second void pixel value correction unit that, in the second M-value image, if there is a region of a different phase from the gas present in the void as a region surrounded by the region corresponding to the void, changes the pixel value of the region of the different phase from the gas included in the second M-value image to the pixel value corresponding to the gas, based on the pixel value of the region of the different phase from the gas in the image to be processed. [Disclosure 13] The processing apparatus according to any one of disclosures 1 to 12, wherein the image processing apparatus has a first phase pixel value correction unit that, when the first M-value image contains a region of a phase different from the single phase as a region surrounded by a single phase, changes the pixel value of the region of the phase different from the single phase included in the first M-value image to the pixel value corresponding to the single phase, based on the size of the region of the phase different from the single phase. [Disclosure 14] The processing apparatus according to any one of disclosures 1 to 13, wherein the image processing apparatus has a second phase pixel value correction unit that, when the second M-value image contains a region of a phase different from the single phase as a region surrounded by a single phase, changes the pixel value of the region of the phase different from the single phase included in the second M-value image to the pixel value corresponding to the single phase, based on the size of the region of the phase different from the single phase. [Disclosure 15] The processing apparatus according to any one of disclosures 1 to 14, wherein the image processing apparatus comprises a feature derivation unit that derives M phase feature quantities based on the second M-valued image. [Disclosure 16] The apparatus according to disclosure 15, wherein the feature quantity includes at least one of the number rate of the M phases, the phase fraction rate, and the aspect ratio. [Disclosure 17] The processing apparatus according to any one of disclosures 1 to 16, wherein M is an integer greater than or equal to 3. [Disclosure 18] The aforementioned substance is a sintered ore, The apparatus according to any one of disclosures 1 to 17, wherein the N phases include hematite, magnetite, and calcium ferrite. [Disclosure 19] An acquisition step involves obtaining an image of a cross-section of a material composed of N phases bonded together, using an optical microscope equipped with an imaging means. An image processing step which includes image processing applied to the captured image, Equipped with, The aforementioned image processing step is: A frequency distribution derivation step for deriving the frequency distribution of pixel values ​​in the image to be processed, A threshold derivation step is performed to derive M-1 thresholds for converting the image to be processed into an M-value image, based on the results of curve fitting the frequency distribution. A first multi-level image generation step generates an image as a first M-level image in which the pixel values ​​of the image to be processed are converted to M-level values ​​based on the M-1 thresholds, A second multi-level image generation step involves identifying regions that form the boundaries of multiple phases in the image to be processed, and generating a second multi-level image by correcting the pixel values ​​of the regions that form the boundaries of multiple phases in the first multi-level image based on the pixel values ​​of the identified regions and the M-1 thresholds. It has, N is an integer greater than or equal to 2, M is an integer between 2 and N, The processing method wherein the image to be processed is the captured image, or an image that has been preprocessed from the captured image. [Disclosure 20] A program for causing a computer to function as a part of the processing apparatus described in any one of disclosures 1 to 18. [Explanation of Symbols]

[0146] 100, 600, 900 processing units 101 Acquisition Department 102, 602, 902 Image Processing Unit 102a Pre-processing unit 102b Frequency distribution derivation part 102c, 902a Threshold derivation section 102d, 902b First multi-level image generation unit 102e Second Multilevel Image Generation Unit 102f Feature Derivation Unit 103 Output section 110 Input Device 120 Output device 210 Unprocessed images 220 Composite images 300 frequency distribution 410 Image to be processed 420 First M-value converted image 430 Image with 2M values 440 Binarized images for phase boundary identification 450 Multilevel images for phase boundary identification 602a First void pixel value correction unit 602b First Phase Intra-Phase Pixel Value Correction Unit 602c Second multilevel image generation unit 602d Second void pixel value correction unit 602e Second Phase Pixel Value Correction Unit 602f Feature Derivation Unit 710 Objects surrounded by voids 1000 Threshold Change Screen 1010 Threshold Derivation Result Display Unit 1020 Threshold reception unit 1021a~1021d Slider 1022a~1022d Threshold display area 1023a~1023d OK button Threshold for generating Th1~Th4M quantified images r1~r5 Pixel value range for multi-level conversion Mean of the Gaussian distribution of μ1-μ5 mixture

Claims

1. An acquisition unit that acquires an image of a cross-section of a material composed of N phases bonded together, using a microscope equipped with an imaging means, An image processing unit that performs image processing on the captured image, Equipped with, The aforementioned image processing unit, A frequency distribution derivation unit that derives the frequency distribution of pixel values ​​in the image to be processed, A threshold derivation unit that derives M-1 thresholds for converting the image to be processed into an M-value image based on the results of curve fitting the frequency distribution, A first multi-level image generation unit generates an image as a first M-level image in which the pixel values ​​of the image to be processed are converted to M-level values ​​based on the M-1 thresholds, A second multi-level image generation unit identifies regions that form the boundaries of multiple phases in the image to be processed, and generates a second multi-level image by correcting the pixel values ​​of the regions that form the boundaries of multiple phases in the first multi-level image based on the pixel values ​​of the identified regions and the M-1 thresholds. It has, N is an integer greater than or equal to 2, M is an integer between 2 and N, The processing apparatus wherein the image to be processed is the captured image, or an image that has been preprocessed from the captured image.

2. The processing apparatus according to claim 1, wherein the image processing unit is an image represented in grayscale.

3. The image processing unit includes a preprocessing unit that performs preprocessing on the captured image, The apparatus according to claim 1 or 2, wherein the preprocessing includes a process for smoothing the captured image.

4. The apparatus according to claim 1 or 2, wherein the preprocessing includes a process for enhancing the edges of the captured image, and a process for combining the pixel values ​​of the captured image without edge enhancement and the pixel values ​​of the captured image with edge enhancement.

5. The processing apparatus according to claim 4, wherein the synthesis includes calculating the average value of the pixel values ​​of the captured image without edge enhancement and the pixel values ​​of the captured image with edge enhancement.

6. The processing apparatus according to claim 1 or 2, wherein the threshold derivation unit derives M-1 thresholds based on the result of curve fitting the frequency distribution with a distribution function obtained by combining M functions, each having one upward-convex peak.

7. The processing apparatus according to claim 1 or 2, wherein the second multi-level image generation unit generates a binarized image by performing adaptive binarization on the image to be processed, and identifies regions that are boundaries between multiple phases in the image to be processed based on the binarized image.

8. The processing apparatus according to claim 1 or 2, comprising: identifying one of M pixel value ranges determined based on M-1 thresholds as the pixel value range corresponding to the region, based on the pixel values ​​of the image to be processed in the region that forms the boundary of the plurality of phases; and collectively changing the pixel values ​​of the region in the first M-value image to the pixel values ​​of the first M-value image obtained by M-value the image to be processed assuming that the pixel values ​​of the image to be processed are within the identified pixel value range.

9. The processing apparatus according to claim 1 or 2, wherein the threshold derivation unit sets at least one of the M-1 thresholds to a threshold specified by the operator.

10. The processing apparatus according to claim 9, wherein the threshold derivation unit, when the result of curve fitting the frequency distribution does not satisfy predetermined conditions, sets at least one of the M-1 thresholds to a threshold specified by the operator.

11. The N phases include gas present in the vacancies of the substance. The processing apparatus according to claim 1 or 2, wherein the image processing apparatus includes a first void pixel value correction unit that, in the first M-value image, if there is a region in the void that is surrounded by a region corresponding to the void and is a region of a different phase from the gas present in the void, changes the pixel value of the region of the different phase from the gas included in the first M-value image to a pixel value corresponding to the gas, based on the pixel value of the region of the different phase from the gas in the image to be processed.

12. The N phases include gas present in the vacancies of the substance. The processing apparatus according to claim 1 or 2, wherein the image processing apparatus has a second void pixel value correction unit that, in the second M-value image, if there is a region of a different phase from the gas present in the void as a region surrounded by the region corresponding to the void, changes the pixel value of the region of the different phase from the gas included in the second M-value image to the pixel value corresponding to the gas, based on the pixel value of the region of the different phase from the gas in the image to be processed.

13. The processing apparatus according to claim 1 or 2, wherein the image processing apparatus has a first phase pixel value correction unit that, when the first M-value image contains a region of a phase different from the single phase as a region surrounded by a single phase, changes the pixel value of the region of the phase different from the single phase included in the first M-value image to the pixel value corresponding to the single phase, based on the size of the region of the phase different from the single phase.

14. The processing apparatus according to claim 1 or 2, wherein the image processing apparatus has a second phase pixel value correction unit that, when the second M-value image contains a region of a phase different from the single phase as a region surrounded by a single phase, changes the pixel value of the region of the phase different from the single phase included in the second M-value image to the pixel value corresponding to the single phase, based on the size of the region of the phase different from the single phase.

15. The processing apparatus according to claim 1 or 2, wherein the image processing apparatus comprises a feature quantity derivation unit that derives M phase feature quantities based on the second M-valued image.

16. The apparatus according to claim 15, wherein the feature quantity includes at least one of the number rate of the M phases, the phase fraction rate, and the aspect ratio.

17. The processing apparatus according to claim 1 or 2, wherein M is an integer of 3 or more.

18. The aforementioned substance is a sintered ore, The apparatus according to claim 1 or 2, wherein the N phases include hematite, magnetite, and calcium ferrite.

19. An acquisition step in which an image is obtained by taking a photograph of a cross-section of a material composed of N phases bonded together using an optical microscope equipped with an imaging means, An image processing step which includes image processing applied to the captured image, Equipped with, The aforementioned image processing step is: A frequency distribution derivation step for deriving the frequency distribution of pixel values ​​in the image to be processed, A threshold derivation step is performed to derive M-1 thresholds for converting the image to be processed into an M-value image, based on the results of curve fitting the frequency distribution. A first multi-level image generation step generates an image as a first M-level image in which the pixel values ​​of the image to be processed are converted to M-level values ​​based on the M-1 thresholds, A second multi-level image generation step involves identifying regions that form the boundaries of multiple phases in the image to be processed, and generating a second multi-level image by correcting the pixel values ​​of the regions that form the boundaries of multiple phases in the first multi-level image based on the pixel values ​​of the identified regions and the M-1 thresholds, It has, N is an integer greater than or equal to 2, M is an integer between 2 and N, The processing method wherein the image to be processed is the captured image, or an image that has been preprocessed from the captured image.

20. A program for causing a computer to function as each part of the processing apparatus according to claim 1 or 2.

Images