JP2016191556A - Image processing method, method of evaluating life of object, and image processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To suppress a reduction in detection accuracy while reducing a time required for detection of a deposit.SOLUTION: The present image processing method is an image processing method of extracting a deposit included in an image of a tissue of a material of an object, ferrite heat-resistant steel. The method includes: an imaging step of photographing the tissue of a material of the object with an electron microscope to create a photographed image; a noise removal step of removing a noise component from the photographed image to create a noise-removed image; a binarization step of performing binarization processing, which is processing of converting an image with a plurality of gradations into an image with two gradations, on the noise-removed image to create a binarized image; and a deposit image extraction step of determining, from an area including one of the gradations of the binarized image, an area in a predetermined shape as an image of a deposit, and removing an area in a shape other than the predetermined shape to extract the image of the deposit.SELECTED DRAWING: Figure 4

Description

  The present invention relates to an image processing method for extracting precipitates from an image of a material structure for performing life evaluation of a ferritic heat resistant steel as an object, an object life evaluation method, and an image processing system.

  For example, ferritic heat-resistant steel used for boilers and the like is used for a long time at a high temperature, and therefore it is necessary to evaluate its life. Life evaluation of ferritic heat resistant steel includes evaluation related to creep life. It has been found that the creep life is affected by the spacing between precipitates contained in the ferritic heat resistant steel. Accordingly, the creep life may be evaluated by observing the metal structure of the ferritic heat resistant steel and measuring the distance between the precipitates. In addition to precipitates, irregularities and scratches on the surface of the metal structure are also observed on the surface of the metal structure of ferritic heat-resistant steel, so it may be difficult to identify the precipitate from the surface of the metal structure. . Therefore, when measuring the distance between the precipitates, an expert generally observes the metal structure and detects the precipitate.

  In order to observe the metal structure of the ferritic heat resistant steel, for example, as shown in Patent Document 1, there is a case in which the metal structure of an object is transferred to a replica and the replica is observed with an electron microscope. Japanese Patent Application Laid-Open No. H10-228561 describes a technique for performing life evaluation by processing an image obtained by imaging a metal structure.

Japanese Patent No. 3015599 Japanese Patent No. 4153675

  However, when an expert performs the detection of the precipitate, since it is a manual analysis, there is a possibility that the detection may take time. In addition, when an operator other than an expert performs detection, for example, the precipitate may be confused with the uneven surface of the metal structure, and the precipitate may not be detected accurately. For example, Patent Document 1 generates a photograph of a metal structure by a replica, but does not describe a device for detecting precipitates. Further, Patent Document 2 describes a system that performs material life evaluation in a short time by image processing, but there is room for improvement in the detection accuracy of precipitates.

  In order to solve the above-described problems, the present invention aims to provide an image processing method, a life evaluation method for an object, and an image processing system that suppress a decrease in detection accuracy while reducing the time required to detect precipitates. And

  In order to solve the above-described problems and achieve the object, the image processing method of the present invention is an image processing method for extracting precipitates contained in an image of a material structure of a ferritic heat-resistant steel as an object, An imaging step of capturing the material structure of the object with an electron microscope to generate a captured image, a noise removing step of generating a noise-removed image by removing a noise component from the captured image, and the noise-removed image A binarization step of generating a binarized image by performing binarization processing that is a process of converting an image having a plurality of gradations into two gradations, and having one gradation of the binarized image Of the regions, a region having a predetermined shape is determined to be an image of the precipitate, and a region other than the predetermined shape is removed to extract the image of the precipitate. Have

  In this image processing method, an image of an area determined not to be a precipitate is removed, and an image of the precipitate is extracted. Further, this image processing method extracts an image of a precipitate according to a predetermined algorithm. Therefore, this image processing method can improve the detection speed while suppressing a decrease in the detection accuracy of precipitates.

  In the image processing method, the noise removal step includes a singular point removal step of generating a singular point removal image by removing a singular point whose gradation is singular with respect to the gradation of surrounding pixels from the captured image. It is preferable that the method further includes a gradation change extraction step of extracting a region having a large gradation change between pixels from the singular point removed image and generating the noise-removed image. Since this image processing method can appropriately remove noise from the captured image, it is possible to more suitably suppress a decrease in the detection accuracy of precipitates.

  In the image processing method, the noise removing step performs a median value filtering process or an average value filtering process in the singular point removing step, and a maximum value filter is applied to the singular point removed image in the gradation change extracting step. It is preferable to generate a low-frequency image by performing processing, and then performing minimum value filtering, and generate the noise-removed image by removing image components of the low-frequency image from the singular point-removed image. . Since this image processing method can appropriately remove noise from the captured image, it is possible to more suitably suppress a decrease in the detection accuracy of precipitates.

  In the image processing method, it is preferable that the noise removal step adds a zero image having a gradation value of zero along the outer periphery of the singular point removal image before the gradation change extraction step. In this image processing method, by adding a zero image, it is possible to suppress the image size from being reduced even when the maximum value filter processing and the minimum value filter processing are performed.

  In the image processing method, in the binarization step, when the pixels in the noise-removed image are divided into high gradation pixels and low gradation pixels, the ratio of the number of high gradation pixels is equal to or higher than a predetermined ratio. A pixel having a gray level as a threshold, a pixel having a gray level equal to or higher than the threshold value as one gray level, and a pixel having a gray level smaller than the threshold value as another gray level P tile processing is preferably performed. With this image processing method, for example, even if the type of image changes, it is possible to suitably suppress a decrease in the detection accuracy of precipitates.

  In the image processing method, it is preferable that the binarization step further performs a maximum value filter process on the noise-removed image subjected to the P tile process, and then performs a minimum value filter process. Since this image processing method can remove noise in the image after the binarization processing, it is possible to suitably suppress a decrease in the detection accuracy of precipitates.

  In the image processing method, the precipitate image extraction step includes a length shorter than a predetermined length, or an area smaller than a predetermined area, or a predetermined area of one gradation region of the binarized image. It is preferable to remove an area having a circularity smaller than the circularity as an image other than the predetermined shape. This image processing method can suitably suppress a decrease in detection accuracy of precipitates.

  In the image processing method, the precipitate image extraction step further includes, in one of the gradation regions of the binarized image, an area smaller than a filling area having an area larger than the predetermined area, and It is preferable to remove an area having a filling rate indicating a ratio of its own area to the area of the circumscribed rectangle smaller than a predetermined value as an image other than the predetermined shape. This image processing method can suitably suppress a decrease in detection accuracy of precipitates.

  In the image processing method, it is preferable that the imaging step obtains the captured image obtained by imaging the replica obtained by transferring the material structure of the object with the electron microscope. According to this image processing method, the life of the object can be evaluated nondestructively.

  It is preferable that the image processing method further includes an inter-precipitate distance calculation step of calculating a distance between the images of the precipitates. According to this image processing method, since the distance between the precipitates is calculated, the life evaluation can be appropriately performed.

  In order to solve the above-described problems and achieve the object, the object life evaluation method of the present invention performs life evaluation of the object based on the precipitate extracted by the image processing method. According to this method for evaluating the lifetime of an object, since an image of the extracted precipitate is used, the lifetime evaluation can be appropriately performed.

  In order to solve the above-described problems and achieve the object, the image processing system of the present invention is an image processing system for extracting precipitates contained in an image of a material structure of a ferritic heat-resistant steel as an object, A captured image acquisition unit that acquires a captured image of a material tissue of the object, a noise removal unit that generates a noise-removed image by removing a noise component from the captured image, and a plurality of gradations with respect to the noise-removed image A binarization processing unit that generates a binarized image by performing a binarization process that is a process of converting an image having two gradations, and among the images of one gradation of the binarized image, And determining a region having a predetermined shape as an image of the precipitate, and removing a region other than the predetermined shape to extract the image of the precipitate. According to this image processing system, it is possible to improve the detection speed while suppressing a decrease in the detection accuracy of precipitates.

  ADVANTAGE OF THE INVENTION According to this invention, the fall of detection accuracy can be suppressed, reducing the time which detection of a deposit requires.

FIG. 1 is a diagram illustrating an example of a metal structure before use of an object. FIG. 2 is a diagram illustrating an example of a metal structure after the object is used. FIG. 3 is a schematic diagram showing the configuration of the life evaluation apparatus according to the present embodiment. FIG. 4 is a block diagram schematically showing the configuration of the control unit. FIG. 5 is an explanatory diagram for describing filter processing on an image. FIG. 6 is an explanatory diagram for explaining the filtering process on an image. FIG. 7 is an explanatory diagram for explaining a zero image. FIG. 8 is a graph illustrating an example for explaining the P tile processing. FIG. 9 is a graph illustrating an example for explaining the P tile processing. FIG. 10 is an explanatory diagram for explaining a precipitate candidate region. FIG. 11 is an explanatory diagram for explaining the precipitate region. FIG. 12 is a diagram for explaining the filling rate. FIG. 13 is a flowchart for describing image processing according to the present embodiment. FIG. 14 is a flowchart illustrating the noise removal process. FIG. 15 is a flowchart for explaining binarization processing. FIG. 16 is a flowchart for explaining the precipitate image extraction process. FIG. 17 is a diagram illustrating an example of a captured image. FIG. 18 is a diagram illustrating an example of an object evaluation image. FIG. 19 is a diagram illustrating an example of an evaluation image according to a comparative example. FIG. 20 is an explanatory diagram illustrating screen transitions when the image processing system according to the present embodiment is operated. FIG. 21 is a diagram illustrating an example of a main screen. FIG. 22 is a diagram illustrating an example of a parameter setting dialog. FIG. 23 is a diagram illustrating an example of a micron bar measurement dialog. FIG. 24 is a diagram illustrating an example of a continuous execution result dialog.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In addition, this invention is not limited by this embodiment, Moreover, when there are two or more embodiments, what comprises a combination of each Example is also included.

FIG. 1 is a diagram illustrating an example of a metal structure before use of an object. FIG. 2 is a diagram illustrating an example of a metal structure after the object is used. Ferritic heat resistant steel is a kind of steel containing additives, and the additives are scattered as precipitates in the metal structure. For example, as shown in FIG. 1, the metal structure of the heat resistant ferritic steel, in steel region S is an area of the steel (ferrite) components of ferritic heat-resistant steel, a plurality of deposit D ₁ are scattered. It has been found that the ferritic heat-resisting steel has a precipitate aggregated and coarsened with use at a high temperature for a long time, and the interval between the precipitates becomes wide. For example, as shown in FIG. 2, when the ferritic heat resistant steel is used for a long time at a high temperature, the precipitate D _{1 in} FIG. 1 aggregates to form a precipitate D ₂ . The distance between the deposit D ₂ is longer than the distance between the deposit D _1. Here, it is known that the distance between the precipitates affects the creep life, and the creep life decreases as the distance between the precipitates increases. Therefore, it is important to detect the distance between precipitates in the life evaluation of ferritic heat resistant steel. The life evaluation system according to the present embodiment performs life evaluation of the object 1 using the ferritic heat resistant steel as the object 1. In addition, although the ferritic heat-resistant steel which is the target object 1 contains chromium as an additive (precipitate), it is not limited thereto, and any component can be used as long as it is a ferritic steel containing a predetermined additive.

  FIG. 3 is a schematic diagram showing the configuration of the life evaluation apparatus according to the present embodiment. As shown in FIG. 3, the life evaluation apparatus 10 includes a control unit 20, a display 22, an input unit 24, and a storage unit 26. The life evaluation apparatus 10 is a computer. The control unit 20 performs computation, and is, for example, a CPU (Central Processing Unit) and a memory. The display 22 is a display that displays a predetermined image under the control of the control unit 20. The input unit 24 includes, for example, a keyboard and a mouse, and is used by an operator to input various information. Information input by the operator includes setting values of parameters related to processing performed by the control unit 20 and instructions for executing the processing. The storage unit 26 is a storage device such as a RAM (Random Access Memory) or a ROM (Read Only Memory). The storage unit 26 stores an image of the material structure of the object 1, parameter setting values related to processing performed by the control unit 20, and the like. In addition, the life evaluation apparatus 10 can acquire data from the electron microscope 170 and store it in the storage unit 26.

  The electron microscope 170 is a scanning electron microscope. When evaluating the life of the object 1, an operator creates a replica 2 to which the material structure of the object 1 is transferred, images the replica 2 with an electron microscope 170, and creates a captured image 100. Specifically, the operator polishes the observation site of the object 1 and corrodes the polished surface by chemical corrosion or electrolytic corrosion. Then, the operator applies the replica 2 to the corroded surface, and transfers the metal structure (irregularities) on the corroded surface to the replica 2. The operator observes the replica 2 to which the metal structure has been transferred with the electron microscope 170 and creates a captured image 100 in which the metal structure of the object 1 transferred to the replica 2 is imaged. The captured image 100 captured by the electron microscope 170 is stored in the storage unit 26 of the lifetime evaluation apparatus 10. In the present embodiment, the life evaluation apparatus 10 is electrically connected to the electron microscope 170, and the captured image 100 is directly input from the electron microscope 170. However, the present invention is not limited to this, and the life evaluation apparatus 10 May not be electrically connected to the electron microscope 170. In this case, the captured image 100 from the electron microscope 170 is stored in the storage unit 26 of the life evaluation apparatus 10 via, for example, an external storage device.

  Next, the control unit 20 will be described in detail. The control unit 20 performs image processing for extracting precipitates included in the material structure image of the object 1. More specifically, the control unit 20 generates an object evaluation image by image processing, and performs a life evaluation of the object 1 based on the object evaluation image. FIG. 4 is a block diagram schematically showing the configuration of the control unit. As shown in FIG. 4, the control unit 20 includes a captured image acquisition unit 32, a parameter acquisition unit 33, a noise removal unit 34, a binarization processing unit 36, a precipitate image extraction unit 37, and an inter-precipitate space. A distance calculation unit 38. The above units of the control unit 20 are software common to each other, but are not limited thereto, and may be software independent of each other, for example. The captured image acquisition unit 32, the parameter acquisition unit 33, the noise removal unit 34, the binarization processing unit 36, the precipitate image extraction unit 37, and the inter-precipitate distance calculation unit 38 include The present invention is included in an image processing system that extracts precipitates contained in a material structure. Further, the image processing system and the life evaluation unit 39 are included in the life evaluation system for the object.

  The captured image acquisition unit 32 acquires the captured image 100 from the storage unit 26. The parameter acquisition unit 33 acquires parameters of processing performed by each unit of the control unit 20 from the storage unit 26.

  The noise removal unit 34 acquires the captured image 100 from the captured image acquisition unit 32, and acquires parameters related to processing executed by the noise removal unit 34 from the parameter acquisition unit 33. The noise removal unit 34 removes noise components from the captured image 100 to generate the noise removal image 102. Specifically, as shown in FIG. 4, the noise removal unit 34 includes a singular point removal unit 42, a zero image addition unit 44, and a gradation change extraction unit 46.

5 and 6 are explanatory diagrams for explaining filter processing on an image. The singularity removal unit 42 performs a median filter (median filter) process on the captured image 100 to generate a singularity removal image 100a. The median filter process is a process of replacing the gradation value (luminance) in the target pixel of the captured image 100 with an intermediate value among the gradation values in the pixels in the peripheral area of the target pixel. Specifically, as illustrated in FIG. 5, the singularity removal unit 42 selects a kernel C ₁ that is a partial region of the captured image 100 including a plurality of pixels from the captured image 100. As shown in FIG. 6, the kernel C ₁ is a region including 3 × 3 pixels having three pixels in the first direction and three pixels in the second direction orthogonal to the first direction. not limited to, the number of pixels included in the kernel C ₁ is arbitrary. The kernel C ₁ has a pixel P _{1 at} the center of the region and pixels P ₂ , P ₃ , P ₄ , P ₅ , P ₆ , P ₇ , P ₈ , P ₉ around the pixel P ₁ . Here, the gradation values (luminances) in the captured image 100 of the pixels P ₁ , P ₂ , P ₃ , P ₄ , P ₅ , P ₆ , P ₇ , P ₈ , P ₉ are set to L ₁ , L ₂ , L, respectively. ₃ , L ₄ , L ₅ , L ₆ , L ₇ , L ₈ , and L ₉ . The singularity removal unit 42 converts the gradation value of the pixel P _{1 from} among L ₁ , L ₂ , L ₃ , L ₄ , L ₅ , L ₆ , L ₇ , L ₈ , and L ₉ by median filtering. (The tone value of the middle (here, the fifth) pixel arranged in descending order). The singularity removal unit 42 performs this median filtering process on all the pixels in the captured image 100. The singularity removal unit 42 performs the median value filter process to smooth the gradation value of a pixel (singularity) having a specific gradation value with respect to the gradation value of the pixel in the peripheral region, The removed singularity removed image 100a is generated.

  Note that the singular point removing unit 42 is not limited to performing the median filter process as long as it removes singular points whose gradation is singular with respect to the gradation of surrounding pixels. The singularity removal unit 42 may perform, for example, average value filtering. The average value filter process is a process of replacing the gradation value (luminance) in the target pixel with the average value of the gradation values in the pixels in the peripheral area of the target pixel.

FIG. 7 is an explanatory diagram for explaining a zero image. The zero image adding unit 44 adds the zero image 101 having a gradation value (luminance) of zero along the outer periphery of the singular point removed image 100a to generate a zero added image 100b. As shown in FIG. 7, the zero image 101 is an image in which pixels are distributed in a frame shape along the outer periphery of the singular point removal image 100a. Each pixel of the zero image 101 has a gradation value of zero. The zero image adding unit 44 adds the zero image 101 to the outer periphery of the singular point removed image 100a, the gradation value of the outermost pixel is zero (the pixel of the zero image), and the inner side is the singular point removed image 100a. A zero addition image 100b is generated. Incidentally, the zero image 101, since one pixel is an image are aligned in a frame shape, as shown in FIG. 7, the width of each side is smaller than the kernel C _1. However, the zero image 101 may have two or more pixels arranged in a frame shape.

  The gradation change extraction unit 46 extracts a region where the gradation change between pixels is large from the zero-added image 100b, and generates the noise-removed image 102. Specifically, the gradation change extraction unit 46 includes a maximum value filter unit 50, a minimum value filter unit 52, and a low frequency image removal unit 54, as shown in FIG.

The maximum value filter unit 50 performs maximum value filtering on the zero-added image 100b. The maximum value filtering process is a process of replacing the gradation value (luminance) in the target pixel of the captured image 100 with the maximum value among the gradation values in the pixels in the peripheral area of the target pixel. Maximum value filter unit 50 selects the kernel C _2, performs the maximum value filtering based on the pixel in the kernel C _2. The kernel C ₂ is an area including 30 × 30 pixels having 30 pixels in the first direction and 30 pixels in the second direction, but is not limited thereto and can be arbitrarily set. Kernel size is necessary to adjust the size that can extract low-frequency components, the kernel C _2, it is preferable to see a relatively large area over a region including the 9 × 9 pixels, for example.

The minimum value filter unit 52 performs a minimum value filter process on the zero-added image 100b that has been subjected to the maximum value filter process, and generates a low-frequency image 100c. The minimum value filtering process is a process of replacing the gradation value (luminance) in the target pixel of the captured image 100 with the minimum value among the gradation values in the pixels in the peripheral area of the target pixel. The minimum value filtering unit 52 selects the kernel C _2, performs the minimum value filtering processing based on the pixel in the kernel C _2.

  As described above, the maximum value filter unit 50 and the minimum value filter unit 52 perform the maximum value filter processing and the minimum value filter processing in this order on the zero-added image 100b, thereby generating the low-frequency image 100c. It can be said that the low-frequency image 100c is an image having a low frequency of change in gradation value between pixels in the image (an image in which change in gradation value is gentle between neighboring pixels). That is, the low-frequency image 100c is an image obtained by removing an area having a large gradation change between pixels from the zero addition image 100b.

  The low frequency image removing unit 54 acquires the zero addition image 100b and the low frequency image 100c. The low frequency image removing unit 54 generates the high frequency image 100d by removing the image component of the low frequency image 100c from the zero addition image 100b. Specifically, the low frequency image removing unit 54 subtracts the gradation value of each pixel of the low frequency image 100c from the gradation value of each pixel of the zero addition image 100b. Thus, the low-frequency image removing unit 54 removes an image component having a low frequency of change in tone value between pixels from the zero-added image 100b, and an image component having a high frequency of change in tone value between pixels ( An image having a large change in gradation value between neighboring pixels) is extracted. That is, the maximum value filter unit 50, the minimum value filter unit 52, and the low-frequency image removal unit 54 generate a high-frequency image 100d obtained by extracting a region having a large gradation change between pixels from the zero-added image 100b. For example, a captured image in which a part of an image (for example, the central part) is bright (or dark) and another part of the image (for example, a peripheral part) is dark (or bright) is acquired according to optical characteristics or lens characteristics. There is a possibility that. By creating a low-frequency image 100c in which the captured image is viewed as a macro, taking a difference between the low-frequency image 100c and the captured image, and generating a high-frequency image 100d, the influence of illumination and microscope characteristics can be removed. The low-frequency image removing unit 54 removes the outermost frame-like image (the gradation value is zero) corresponding to the zero image 101 from the high-frequency image 100d to generate the noise-removed image 102.

  As long as the gradation change extraction unit 46 extracts a region where the gradation change between pixels is large, the maximum value filter unit 50, the minimum value filter unit 52, and the low-frequency image removal unit as described above. 54 is not limited thereto.

  As described above, the noise removing unit 34 removes the noise component from the captured image 100 to generate the noise removed image 102. In addition, the noise removal part 34 does not have the zero image addition part 44, but the gradation change extraction part 46 may produce | generate the noise removal image 102 directly from the singular point removal image 100a.

  The binarization processing unit 36 acquires the noise-removed image 102 from the noise removing unit 34, and acquires parameters related to the process executed by itself from the parameter acquiring unit 33. The binarization processing unit 36 performs a binarization process, which is a process of converting an image composed of a plurality of gradations into an image composed of two gradations, on the noise-removed image 102, and uses only two kinds of gradations. A binarized image 104 that is an image to be generated is generated. Specifically, the binarization processing unit 36 includes a P tile unit 60, a minimum value filter unit 62, and a maximum value filter unit 64, as shown in FIG.

  The P tile unit 60 performs P tile processing as binarization processing on the noise-removed image 102 to generate a P tile image 102a. 8 and 9 are graphs showing an example for explaining the P tile processing. In the P tile processing, when each pixel in the noise-removed image 102 is divided into two, a high gradation pixel and a low gradation pixel, a gradation value in which the ratio of the number of high gradation pixels is equal to or higher than a predetermined ratio is obtained. This is a process of setting a threshold value, a gradation value of a pixel having a gradation value equal to or higher than the threshold value as a first gradation, and a gradation value of a pixel having a gradation value smaller than the threshold value as a second gradation.

  An example of P tile processing will be described with reference to FIGS. The horizontal axis of FIG. 8 is the gradation value (luminance) of each pixel in a predetermined image, and the vertical axis is the number of pixels for each gradation value. The horizontal axis of FIG. 9 is the gradation value (luminance) of each pixel in a predetermined image, and the vertical axis is the ratio of the total number of pixels for each gradation value to the total number of pixels (the integrated value of the number of pixels). is there. When the gradation value of each pixel in the image has the distribution shown in FIG. 8, the cumulative number of pixels is as shown in FIG. In other words, the total number of pixels in the predetermined gradation value shown in FIG. 9 is the ratio of the total number of pixels having gradation values equal to or lower than the predetermined gradation value to the whole. The P tile unit 60 detects the gradation value B when the integrated pixel number is a predetermined integrated pixel number A (%), and sets the gradation value B as a threshold value. Note that the value of the cumulative pixel number A (%) is preset by parameter setting. The P tile unit 60 sets the gradation value of a pixel having a gradation value equal to or higher than the gradation value B in the noise-removed image 102 as the first gradation, and the gradation of the pixel having a gradation value smaller than the gradation value B. The P tile image 102a is generated with the value as the second gradation. The P tile image 102a is an image having only two types of gradations, a first gradation and a second gradation. The first gradation has a larger gradation value than the second gradation. The first gradation is, for example, a value that maximizes the luminance, and the second gradation is a value that minimizes the luminance. However, the first gradation and the second gradation can be arbitrarily set as long as they have different values.

The minimum value filter unit 62 performs minimum value filtering on the P tile image 102a. The minimum value filtering unit 62 selects the kernel C _3, performs the minimum value filtering processing based on the pixel in the kernel C _3. The maximum value filter unit 64 performs maximum value filtering on the P tile image 102a that has been subjected to minimum value filtering. Maximum value filter unit 64 selects the kernel C _3, performs the maximum value filtering based on the pixel in the kernel C _3, and generates a binarized image 104.

The minimum value filter unit 62 and the maximum value filter unit 64 perform the minimum value filter processing and the maximum value filter processing on the P tile image 102a in this order, thereby removing a region where the gradation change between adjacent pixels is large. Images are generated and shapes between pixels having different gradation values are shaped. The kernel C ₃ is an area including 3 × 3 pixels having three pixels in the first direction and three pixels in the second direction, but is not limited thereto and can be arbitrarily set.

  The precipitate image extraction unit 37 acquires the binarized image 104 from the binarization processing unit 36, and acquires parameters related to processing executed by the parameter acquisition unit 33. The precipitate image extraction unit 37 determines that an image having a predetermined shape is an image of the precipitate D among images of one gradation of the binarized image 104, and removes an image other than the predetermined shape. Thus, the object evaluation image 106 from which the image of the precipitate D is extracted is generated. The precipitate D is a precipitate scattered in the steel region S of the object 1. Specifically, the precipitate image extraction unit 37 includes a precipitate candidate image extraction unit 66 and an other component image removal unit 68.

The precipitate candidate image extraction unit 66 extracts images that are candidates for the image of the precipitate D from the binarized image 104. The precipitate candidate image extraction unit 66 selects, as the precipitate candidate region X, an area composed of pixels having the first gradation in the binarized image 104. The precipitate candidate region X is a region estimated by the precipitate candidate image extraction unit 66 as an image of the precipitate D. FIG. 10 is an explanatory diagram for explaining a precipitate candidate region. As shown in FIG. 10, in the precipitate candidate area X, a plurality of areas X ₁ including at least one pixel are scattered in the binarized image 104. FIG. 10 is an example showing the precipitate candidate region X.

The other component image removing unit 68 determines that the region X ₁ having a predetermined shape among the regions X ₁ included in the precipitate candidate region X is an image of the precipitate D. Then, the other component image removing unit 68 determines that the region X ₁ having no predetermined shape from the precipitate candidate region X is not an image of the precipitate D and removes it. Region X ₁ that remains without being removed, forms a region which is determined as an image which precipitates region Y of the deposit D. Other ingredients image removing unit 68, by removing the region X ₁ from precipitation candidate region X having no predetermined shape to produce an object image for evaluation 106 obtained by extracting the precipitate region Y. Specifically, the other component image removing unit 68, the tone value of area X ₁ without a predetermined shape of the precipitate candidate region X, replacing the first gradation into the second gradation. Accordingly, the object evaluation image 106 is an image in which the precipitate region Y has the first gradation and the other regions have the second gradation. FIG. 11 is an explanatory diagram for explaining the precipitate region. FIG. 11 shows a state where the precipitate region Y is extracted from the precipitate candidate region X shown in FIG. The precipitate region Y in the object evaluation image 106 is left without removing the region X ₁ having a predetermined shape from the region X _{1 in} the precipitate candidate region X.

Here, the predetermined shape of the region X ₁ to be determined as a precipitate area Y will be described. The other component image removal unit 68 uses the size and shape of the region as elements for determining a predetermined shape. Other ingredients image removing unit 68, when the region X ₁ is smaller than a predetermined size, it is determined not to be a precipitation region Y. Also, another component image removal unit 68, when the shape of the region X ₁ is complex, it is determined not to be a precipitation region Y. Other ingredients image removing unit 68, for example, if the perimeter of the area X ₁ to the area of a region X ₁ is a predetermined value or more, as the shape of the area is complicated, the area X ₁ precipitates region It is determined that it is not Y.

Specifically, the other component image removing unit 68 uses the length in the first direction, the length in the second direction, the area, the circularity, and the filling rate. Specifically, the other component image removing unit 68, removed the length in the first direction a short region X ₁ than the predetermined length J1, determines that not the precipitates region Y (not the image of the deposit D) To do. Also, another component image removing section 68, the length in the second direction a short region X ₁ than the predetermined length J2, it is determined to remove is not the precipitates region Y. Also, other ingredients image removing unit 68, area a predetermined area J3 smaller area X _1, it is determined to remove is not the precipitates region Y. Also, other ingredients image removing unit 68, circularity a predetermined circularity J4 smaller area X _1, it is determined not to be a deposit area Y is removed. The circularity is an index indicating the complexity of the shape of the outer periphery, and is calculated by the following formula (1) when the circularity is Ro.

Ro = (4 · π · A _r ) / L _e ² (1)

Here, π is pi, A _r is the area of region X _1, L _e is the perimeter of the area X _1. The circularity Ro becomes a shape that moves away from a perfect circle as the value decreases from 1.

Also, other ingredients image removing unit 68, an area is smaller than the predetermined area J5, and the filling rate is a predetermined filling rate J6 smaller area X _1, it is determined to remove is not the precipitates region Y. The predetermined area J5 is larger than the predetermined area J3. FIG. 12 is a diagram for explaining the filling rate. The filling rate is a value indicating the area of the region K1 with respect to the area of the circumscribed rectangle K2 circumscribing the region K1, as shown in FIG. That is, the filling rate M is expressed by the following formula (2).

M = Ar _K1 / Ar _K2 (2)

Here, Ar _K1 is the area of the region K1, and Ar _K2 is the area of the circumscribed rectangle K2.

Other ingredients image removing unit 68 by extracting the above manner, precipitate candidate region X from deposit (D) regions X ₁ precipitation region Y by removing the determined not to be the image of the object evaluation A working image 106 is generated.

The inter-precipitate distance calculation unit 38 acquires the object evaluation image 106 from the precipitate image extraction unit 37, and acquires parameters related to processing executed by itself from the parameter acquisition unit 33. The inter-precipitate distance calculation unit 38 calculates the distance between the images of the precipitate D in the object evaluation image 106. Specifically, the inter-precipitate distance calculation unit 38 measures the distance between the regions X ₁ included in the precipitate region Y from the object evaluation image 106. It precipitates distance calculation unit 38, as the distance between the regions X _1, calculates the mean free path between area X _1. The inter-precipitate distance calculation unit 38 may add a measurement line for calculating the distance between the precipitates D to the object evaluation image 106.

The life evaluation unit 39 acquires information on the distance between the regions X ₁ from the inter-precipitate distance calculation unit 38. Life evaluating section 39, based on the distance information between the regions X _1, calculates the lifetime of the object 1. As described above, the distance between the precipitates D is related to the creep life of the object 1. The longer the distance between the precipitates D, the shorter the creep life. Since the region X ₁ included in the precipitate region Y shows an image of the precipitate D, the life evaluation unit 39 calculates the creep life with the distance between the regions X _{1 as} the distance between the precipitates D. .

  The processing flow of the control part 20 demonstrated above is demonstrated based on a flowchart. FIG. 13 is a flowchart for describing image processing according to the present embodiment. As illustrated in FIG. 13, when performing the image processing according to the present embodiment, the control unit 20 acquires the captured image 100 of the target object 1 using the captured image acquisition unit 32 (step S10), and the parameter acquisition unit 33 Various parameters necessary for image processing are acquired (step S12).

  After acquiring the captured image 100 and parameters, the control unit 20 performs noise removal processing by the noise removal unit 34 (step S14) to generate the noise removal image 102. A detailed processing flow of the noise removal processing will be described later. After executing the noise removal processing, the control unit 20 executes the binarization processing by the binarization processing unit 36 (step S16) and generates the binarized image 104. A detailed processing flow of the binarization processing will be described later. After executing the binarization process, the control unit 20 causes the precipitate image extraction unit 37 to execute a precipitate image extraction process (step S18), and generates an object evaluation image 106. A detailed processing flow of the precipitate image extraction processing will be described later.

After executing the precipitate image extraction process, the control unit 20 causes the precipitate distance calculation unit 38 to execute a precipitate distance calculation process from the object evaluation image 106 (step S19). Specifically, it precipitates distance calculation unit 38, by measuring the distance between the regions X ₁ contained in the precipitate region Y of the object in the image for evaluation 106 calculates the distance between the deposit (D) . Thereby, the image processing according to the present embodiment ends. Thereafter, the control unit 20 calculates the creep life of the object 1 by the life evaluation unit 39 based on the calculated distance information between the precipitates D.

  Next, the processing flow of noise removal processing will be described. FIG. 14 is a flowchart illustrating the noise removal process. As shown in FIG. 14, when performing the noise removal process, the noise removal unit 34 performs the singular point removal process on the captured image 100 by the singular point removal unit 42 (step S20), and the singular point removal image 100a is obtained. Generate. The singularity removal unit 42 performs median filter processing as singularity removal processing. After performing the singularity removal processing, the noise removal unit 34 adds the zero image 101 to the singularity removal image 100a by the zero image addition unit 44 (step S22), and generates the zero addition image 100b. After adding the zero image 101, the noise removing unit 34 performs the maximum value filtering process and the minimum value filtering process on the zero added image 100b by the maximum value filter unit 50 and the minimum value filter unit 52, and generates the low frequency image 100c. (Step S24). The maximum value filter unit 50 and the minimum value filter unit 52 execute the maximum value filter processing and the minimum value filter processing in this order.

  After generating the low frequency image 100c, the noise removal unit 34 removes the image component of the low frequency image 100c from the singular point removal image 100a by the low frequency image removal unit 54 (step S26), and generates the noise removal image 102. To do. Thus, the noise removal process is finished.

  Next, the processing flow of binarization processing will be described. FIG. 15 is a flowchart for explaining binarization processing. As shown in FIG. 15, when performing binarization processing, the binarization processing unit 36 performs P tile processing on the noise-removed image 102 by the P tile unit 60 (step S30), and the P tile image 102a. Is generated. After performing the P tile processing, the binarization processing unit 36 performs the minimum value filtering process on the P tile image 102a by the minimum value filter unit 62 (step S32), and then the maximum value filter unit 64 performs the maximum value processing. A value filter process is executed (step S34), and a binary image 104 is generated. Thereby, the binarization process ends.

Next, the processing flow of the precipitate image extraction process will be described. FIG. 16 is a flowchart for explaining the precipitate image extraction process. As shown in FIG. 16, when performing the precipitate image extraction process, the precipitate image extraction unit 37 selects the precipitate candidate region X from the binarized image 104 by the precipitate candidate image extraction unit 66 (step S40). The precipitate candidate image extraction unit 66 selects, as the precipitate candidate region X, an area composed of pixels having the first gradation in the binarized image 104. After selecting the precipitate candidate region X, the precipitate image extracting unit 37 extracts the precipitate region Y from the precipitate candidate region X by the other component image removing unit 68 (step S42), and the object evaluation image. 106 is generated. Other ingredients image removing unit 68, among the precipitates candidate area region X ₁ in X, to remove the region X ₁ having no predetermined shape, the image of the deposit (D) a region X ₁ having a predetermined shape It is judged that there is, and is extracted as a precipitate region Y. Thereby, the precipitate image extraction process ends.

  An example of the object evaluation image 106 generated by the image processing system according to this embodiment is shown below. FIG. 17 is a diagram illustrating an example of a captured image. FIG. 18 is a diagram illustrating an example of an object evaluation image. FIG. 19 is a diagram illustrating an example of an evaluation image according to a comparative example. FIG. 17 is a captured image 100 </ b> A that is an example of the captured image 100 of the object 1, and precipitates D are scattered in the steel region S. FIG. 18 shows an object evaluation image 106A generated by the image processing system according to this embodiment based on the captured image 100A of FIG. As shown in FIG. 18, in the object evaluation image 106A, the image of the precipitate D has the first gradation as the precipitate area Y, and the other area has the second gradation as the steel area S. Yes. In FIG. 18, the distance between the precipitates D is calculated by the inter-precipitate distance calculation unit 38 for the object evaluation image 106 </ b> A, and a measurement line for calculating the distance between the precipitates D is shown. It has been added to the image. FIG. 19 shows an evaluation image 106X in which an operator detects an image of the precipitate D based on the captured image 100A of FIG. 17 and extracts the detected image of the precipitate D. The evaluation image 106X is created by an operator and is an image according to a comparative example. The evaluation image 106X shown in FIG. 19 is also an image in which the precipitates D are scattered in the steel region S, similarly to the object evaluation image 106A shown in FIG. 18 and 19, it can be seen that the object evaluation image 106A has the same distribution of precipitates D as the evaluation image 106X extracted by the operator. Also from these figures, it can be seen that the image processing system according to the present embodiment suppresses a decrease in the detection accuracy of the precipitate D.

  Table 1 shows the calculation result of the distance between precipitates calculated based on the object evaluation image according to the present embodiment, and the calculation result of the distance between precipitates calculated based on the evaluation image according to the comparative example. ing. The distance between precipitates in Table 1 is indicated by the mean free path, and is in units of μm and the number of pixels. Note that the relationship between the number of pixels and μm changes according to the magnification of the image.

  As shown in Table 1, the calculation result of the distance between precipitates based on the sample 1 of the captured image 100 is 1.53 μm (211 pixels) in the present embodiment, and 1.93 μm (266 pixels) in the comparative example. Met. Moreover, the calculation result of the distance between precipitates based on the sample 2 of the captured image 100 was 2.23 μm (307 pixels) in the present embodiment, and 2.37 μm (327 pixels) in the comparative example. In addition, the calculation result of the distance between the precipitates based on the sample 3 of the captured image 100 is 1.94 μm (268 pixels) in the present embodiment, and 1.92 μm (264 pixels) in the comparative example. Further, the calculation result of the distance between the precipitates based on the sample 4 of the captured image 100 is 1.79 μm (247 pixels) in the present embodiment, and 1.75 μm (241 pixels) in the comparative example. Moreover, the calculation result of the distance between the precipitates based on the sample 5 of the captured image 100 is 2.20 μm (the number of 303 pixels) in the present embodiment, and 2.66 μm (the number of 367 pixels) in the comparative example.

  As shown in Table 1, the calculation result of the inter-precipitate distance calculated based on the object evaluation image according to the present embodiment has a small change from the comparative example, and suppresses a decrease in the detection accuracy of the precipitate D. I understand that.

As described above, the image processing system configured by the control unit 20 according to the present embodiment is an image processing system that extracts the precipitate D included in the image of the material structure of the object 1. The image processing system according to the present embodiment includes a captured image acquisition unit 32, a noise removal unit 34, a binarization processing unit 36, and a precipitate image extraction unit 37. The captured image acquisition unit 32 acquires a captured image 100 of the material tissue of the target 1. The noise removing unit 34 removes noise components from the captured image 100 to generate a noise removed image 102. The binarization processing unit 36 performs binarization processing on the noise-removed image 102 and generates a binarized image 104. Precipitates image extracting section 37, binarization of one gradation of the image of the image 104, it is determined that the image of the deposit (D) a region X ₁ is a predetermined shape, a region other than a predetermined shape by removing X _1, extracts the image of the deposit D.

  According to the image processing system according to the present embodiment, the image of the material structure of the object 1 is binarized to select a region that is a candidate for the precipitate D (precipitate candidate region X), and the precipitate From the candidate region X, the image of the region determined not to be the precipitate D is removed, and the image of the precipitate D is extracted. Since the image processing system according to the present embodiment extracts the image of the precipitate D according to a predetermined algorithm, for example, the operator can detect the precipitate D rather than manually extracting the image of the precipitate D. Speed can be improved. In addition, the image processing system according to the present embodiment performs processing of removing noise from the captured image 100, performing binarization processing thereafter, and then extracting an image determined to be the precipitate D. A decrease in the detection accuracy of D can be suppressed. In the present embodiment, the object evaluation image 106 is generated based on the captured image 100, but if the image of the precipitate D is extracted based on the captured image 100, the object evaluation image 106 is generated. You don't have to.

  The noise removing unit 34 includes a singular point removing unit 42 and a gradation change extracting unit 46. The singular point removal unit 42 removes singular points whose gradation is singular with respect to the gradation of surrounding pixels from the captured image 100 to generate a singular point removed image 100a. The gradation change extraction unit 46 extracts a region where the gradation change between pixels is large from the singular point removal image 100 a to generate the noise removal image 102. The noise removing unit 34 removes singular points and then extracts a region having a large gradation change between pixels, so that noise can be removed appropriately from the captured image 100. Therefore, this image processing system can more suitably suppress a decrease in the detection accuracy of the precipitate D.

  Further, the singularity removal unit 42 performs median value filtering processing or average value filtering processing, and the gradation change extraction unit 46 performs maximum value filtering processing on the singularity removal image 100a, and then performs minimum value filtering processing. By performing, the low frequency image 100c is produced | generated and the noise removal image 102 is produced | generated by removing the image component of the low frequency image 100c from the singular point removal image 100a. This image processing system generates a low-frequency image 100c after performing median filtering or average filtering, and performs processing to remove image components from the original image (singular point removed image 100a). Therefore, this image processing system can appropriately remove noise from the captured image 100 and more appropriately suppress a decrease in the detection accuracy of the precipitate D.

  Further, the noise removing unit 34 adds a zero image along the outer periphery of the singular point removed image 100 a by the zero image adding unit 44 before the processing of the gradation change extracting unit 46. The zero image adding unit 44 adds a zero image to the outer periphery of the image before the maximum value filter and minimum value filter processing by the gradation change extracting unit 46. By adding the zero image, it is possible to prevent the image size from being reduced due to the invalid area on the outermost periphery during the maximum value filter processing and the minimum value filter processing.

  Also, the binarization processing unit 36 performs P tile processing by the P tile unit 60. Since the binarization processing unit 36 can determine the threshold value of the gradation in the binarization process by the P tile process, for example, even if the type of the image changes, the decrease in the detection accuracy of the precipitate D is suitably suppressed. be able to.

  In addition, the binarization processing unit 36 performs the minimum value filtering process on the image that has been subjected to the P tile processing by the minimum value filtering unit 62 and the maximum value filtering unit 64, and then performs the maximum value filtering process. Since the binarization processing unit 36 can remove noise in the image after the binarization processing by this processing, it is possible to suitably suppress a decrease in the detection accuracy of the precipitate D.

  Further, the precipitate image extraction unit 37 has a length shorter than the predetermined length J1 (or J2) in one gradation region of the binarized image 104, or an area smaller than the predetermined area J3, or An area having a circularity smaller than the predetermined circularity J4 is removed as an image other than the predetermined shape. In general, the precipitate D is not less than a certain size and the shape is not too complicated. Since the precipitate image extraction unit 37 can remove a region having a size that is too small or a region having a complicated shape as not being the precipitate unit D, it appropriately suppresses a decrease in the detection accuracy of the precipitate D. be able to.

  In addition, the precipitate image extraction unit 37 has a region whose area is smaller than the predetermined area J5 (filling area) and whose filling rate is smaller than the predetermined value in one gradation area of the binarized image 104. Then, the image is removed as an image having a shape other than the predetermined shape. Since this deposit image extraction unit 37 can remove a region having a small filling rate and an excessively complicated shape as not being the precipitate portion D, it is possible to suitably suppress a decrease in the detection accuracy of the precipitate D. it can.

  In addition, the captured image acquisition unit 32 acquires a captured image 100 obtained by capturing an image of the replica 2 to which the material structure of the target 1 is transferred with the electron microscope 170. Since the captured image 100 is an image of the replica 2, the life evaluation of the object 1 can be performed nondestructively.

  Further, the image processing system according to the present embodiment includes an inter-precipitate distance calculation unit 38 that calculates the distance between the images of the precipitate D. Since this image processing system calculates the distance between the precipitates D by the inter-precipitate distance calculation unit 38, the life evaluation can be appropriately performed.

  Hereinafter, a screen configuration in the display 22 when the image processing system according to the present embodiment is operated by the life evaluation apparatus 10 will be described. FIG. 20 is an explanatory diagram illustrating screen transitions when the image processing system according to the present embodiment is operated. When the image processing system is operated, each operation screen is displayed on the display 22.

  Specifically, as shown in FIG. 20, when the operator activates the system by the input unit 24, the main screen 110 is displayed on the display 22. When the operator decides to select a file from the main screen 110, a file selection dialog 120 is displayed. Here, the file is a file in which the captured image 100 for generating the object evaluation image 106 is stored, and the file in which the captured image 100 is stored is displayed in the file selection dialog 120. . When the operator selects a file of the captured image 100 to be processed from the file selection dialog 120, the screen returns to the main screen 110. Here, the operator can simultaneously select a plurality of captured images 100 and continuously execute processing on the plurality of captured images 100. Even if the operator cancels the file selection, the screen returns to the main screen 110.

  As shown in FIG. 20, on the main screen 110 after file selection, when the operator decides to set parameters, a parameter setting dialog 130 is displayed. The parameter setting dialog 130 is a screen for setting various parameters for generating the object evaluation image 106. When the operator completes the parameter setting from the parameter setting dialog 130, the screen returns to the main screen. Even if the operator cancels the parameter setting, the screen returns to the main screen 110.

  Further, as shown in FIG. 20, when the operator decides to measure the length of each part of the image with the micron bar in the parameter setting dialog 130, the micron bar measurement dialog 140 is displayed. The micron bar measurement dialog 140 is a screen for displaying a micron bar in the captured image 100 and adjusting the length of the micron bar in the image. A micron bar is a bar indicating a reference dimension in an image. When the operator completes various settings of the micron bar from the micron bar measurement dialog 140, the process returns to the parameter setting dialog 130. Even if the operator cancels various settings of the micron bar, the parameter setting dialog 130 is displayed again. The micron bar setting may or may not be performed.

  As shown in FIG. 20, on the main screen 110 after parameter setting, when the operator selects to continuously generate the object evaluation image 106, the object evaluation image 106 is created, and the continuous execution result dialog is displayed. 150 is displayed. Continuous execution is to continuously execute processing on a plurality of captured images 100, and the continuous execution result dialog 150 displays information on the distance between the precipitates D based on the respective object evaluation images 106. The When the operator finishes the continuous execution process in the continuous execution result dialog 150, the screen returns to the main screen 110.

  Details of an example of each screen will be described below. FIG. 21 is a diagram illustrating an example of a main screen. As shown in FIG. 21, the main screen 110 displays displays 111, 112, 113, 114, 117 and process command units 116, 118. The display 111 displays the captured image 100. The display 112 displays an image in which a measurement line for calculating the distance between the precipitates D is added to the object evaluation image 106 generated based on the captured image 100 of the display 111. The display 113 displays an image obtained by superimposing the object evaluation image 106 on the captured image 100. The display 114 displays information on the distance between the precipitates D based on the object evaluation image 106 corresponding to the displays 111, 112, and 113. The display 117 displays the size of the image displayed on the displays 111, 112, 113, and 114. The displays 111, 112, 113, and 114 are displayed on the same screen, so that respective images can be compared.

  The processing command unit 116 has an end command for terminating the system, a file selection command, a parameter setting command, a continuous execution command, a continuous binarization command, an editing operation mode selection command, and the like. By selecting each command, These processes can be performed. The processing command unit 118 is a command used when an edit operation mode selection command is selected, and is a command for manually editing the captured image 100 or the object evaluation image 106.

  FIG. 22 is a diagram illustrating an example of a parameter setting dialog. As shown in FIG. 22, the parameter setting dialog 130 displays displays 131, 132, 133, 134, 136, and a processing command unit 135. The display 131 is a display in which a kernel size parameter in each filter process can be set by inputting a numerical value to the parameter input unit 131a. The display 132 is a display for setting parameters in the P tile process by inputting a numerical value to the parameter input unit 132a. The display 133 is a display for setting a parameter in the precipitate image extraction, and a threshold value for determining an area to be removed as not the precipitate area Y is set by inputting a numerical value to the parameter input unit 133a. Can do. The display 134 can set the distance calculation method between the precipitates D by inputting a numerical value to the parameter input unit 134a. The display 136 displays the ratio between the reference length and the number of pixels in the displayed image. The processing command unit 135 includes a micron bar measurement command, a command for reading parameters from a file, a command for saving parameters in a file, and the like.

  FIG. 23 is a diagram illustrating an example of a micron bar measurement dialog. As shown in FIG. 23, in the micron bar measurement dialog 140, a display 141 and processing commands 144, 145, and 146 are displayed. The display 141 displays a captured image 100 and a micron bar 142. The processing command 144 is a command for selecting an image displayed on the display 141. The processing command 145 is a command for setting the length of the micron bar 142. The processing command 146 is a command for setting and canceling the micron bar 142. The micron bar 142 includes end portions 142a and 142b and a line portion 142c between the end portions 142a and 142b. Here, the captured image 100 is an image enlarged at a predetermined magnification. The micron bar 142 is a bar indicating a reference dimension for indicating an actual dimension of the captured image 100. In the micron bar measurement dialog 140, by changing the setting with the processing command 145, it is possible to change the relationship between the reference dimension of the cron bar 142 and the number of pixels in the captured image 100, that is, the magnification of the image. In the micron bar measurement dialog 140, the magnification of the image can also be changed by adjusting the positions of the ends 142a and 142b to change the relative length of the micron bar 142 with respect to the captured image 100. Yes. In this case, the control unit 20 can also change the parameters set in the parameter setting dialog 130 in conjunction with the change in the image magnification. That is, when the magnification is changed to 2 times, the control unit 20 sets the predetermined length, which is a parameter used when determining the precipitate region Y, to be 2 times before the magnification change, and changes the predetermined area to the magnification. 4 times the previous time.

  FIG. 24 is a diagram illustrating an example of a continuous execution result dialog. As shown in FIG. 24, the continuous execution result dialog 150 displays displays 151, 152, 153, and a processing command section 154. In the displays 151 and 152, processing parameters and the like are displayed. The display 153 displays a plurality of information on the distance between the precipitates D based on the object evaluation image 106. The processing command unit 154 has a command for ending this processing and a command for printing the result.

  As mentioned above, although embodiment of this invention was described, embodiment is not limited by the content of this embodiment. In addition, the above-described constituent elements include those that can be easily assumed by those skilled in the art, those that are substantially the same, and those in a so-called equivalent range. Furthermore, the above-described components can be appropriately combined. Furthermore, various omissions, substitutions, or changes of the components can be made without departing from the spirit of the above-described embodiment.

DESCRIPTION OF SYMBOLS 1 Object 2 Replica 10 Life evaluation apparatus 20 Control part 22 Display 24 Input part 26 Storage part 32 Captured image acquisition part 33 Parameter acquisition part 34 Noise removal part 36 Binarization process part 37 Precipitate image extraction part 38 Distance between precipitates Calculation unit 39 Life evaluation unit 42 Singular point removal unit 44 Zero image addition unit 46 Gradation change extraction unit 50, 64 Maximum value filter unit 52, 62 Minimum value filter unit 54 Low frequency image removal unit 60 P tile unit 66 Precipitate candidate Image extraction unit 68 Other component image removal unit 100 Captured image 100a Singularity removal image 100b Zero added image 100c Low frequency image 100d High frequency image 101 Zero image 102 Noise removal image 102a P tile image 104 Binary image 106 Image for object evaluation 170 Electron microscope D Precipitate S Steel region X Precipitate candidate region Area X ₁ area Y Precipitate area

Claims (12)

An image processing method for extracting precipitates contained in an image of a material structure of a ferritic heat resistant steel that is an object,
An imaging step of imaging the material structure of the object with an electron microscope to generate a captured image;
A noise removing step of removing a noise component from the captured image to generate a noise-removed image;
A binarization step of generating a binarized image by performing a binarization process that is a process of converting an image having a plurality of gradations into two gradations on the noise-removed image;
Of the region having one gradation of the binarized image, a region having a predetermined shape is determined to be an image of the precipitate, and by removing the region other than the predetermined shape, A precipitate image extracting step for extracting an image; and
An image processing method. The noise removing step includes:
A singular point removal step of generating a singular point removal image by removing singular points whose gradation is singular with respect to the gradation of surrounding pixels from the captured image;
2. The image processing method according to claim 1, further comprising: a gradation change extraction step of extracting a region having a large gradation change between pixels from the singular point-removed image and generating the noise-removed image. The noise removing step includes:
In the singularity removal step, median filtering or average filtering is performed,
In the gradation change extracting step, a maximum value filtering process is performed on the singular point removal image, and then a minimum value filtering process is performed to generate a low frequency image, and the low frequency image is generated from the singular point removal image. The image processing method according to claim 2, wherein the noise-removed image is generated by removing the image component.   The noise removal step adds a zero image having a gradation value of zero along the outer periphery of the singular point removal image before the gradation change extraction step. Image processing method.   In the binarization step, when pixels in the noise-removed image are divided into high gradation pixels and low gradation pixels, a gradation at which the ratio of the number of high gradation pixels is a predetermined ratio or more is set as a threshold value. P tile processing in which a pixel having a gradation equal to or higher than the threshold is set as one of the two gradations, and a pixel having a gradation smaller than the threshold is set as another of the two gradations. The image processing method according to claim 1, wherein the image processing method is performed.   6. The image processing method according to claim 5, wherein the binarization step further performs a maximum value filter process on the noise-removed image subjected to the P tile process, and then performs a minimum value filter process.   In the precipitate image extraction step, the length of one gradation region of the binarized image is shorter than a predetermined length, or an area is smaller than a predetermined area, or the circularity is higher than a predetermined circularity. The image processing method according to any one of claims 1 to 6, wherein a region having a small size is removed as an image having a shape other than the predetermined shape.   The precipitate image extraction step further includes a region having a smaller area than a filling area having an area larger than the predetermined area in one gradation region of the binarized image, and the area of a circumscribed rectangle of the region. The image processing method according to claim 7, wherein an area having a filling rate indicating a ratio of the area is smaller than a predetermined value is removed as an image other than the predetermined shape.   9. The image processing method according to claim 1, wherein the imaging step acquires the captured image obtained by imaging the replica obtained by transferring the material structure of the object with the electron microscope.   Furthermore, the image processing method of any one of Claim 1 to 9 which has the distance calculation step between precipitates which calculates the distance between the images of the said precipitate.   The life evaluation method of the target object which performs the life evaluation of the said target object based on the said deposit extracted by the image processing method of any one of Claims 1-10. An image processing system for extracting precipitates contained in an image of a material structure of a ferritic heat resistant steel that is an object,
A captured image acquisition unit that acquires a captured image of the material tissue of the object;
A noise removing unit that removes a noise component from the captured image to generate a noise-removed image;
A binarization processing unit that generates a binarized image by performing binarization processing that is a process of converting an image having a plurality of gradations into two gradations on the noise-removed image;
By determining that a region having a predetermined shape is an image of the precipitate in one gradation image of the binarized image, and removing the region other than the predetermined shape, the precipitate A precipitate image extraction unit for extracting the image of
An image processing system.