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

Abstract

A control device of the present invention comprises a CPU. The CPU receives the position of a target object in a captured image, sets region-of-interest candidates which are candidates for a region of interest from the captured image on the basis of the position of the target object, calculates, for each region-of-interest candidate, an image quality evaluation value which is an index value for evaluating the image quality of the captured image, and selects the region of interest of the captured image from among the region-of-interest candidates on the basis of the image quality evaluation value.

Description

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

[0001] This disclosure relates to an image processing apparatus, an image processing method, and an image processing program.

[0002] Industrial parts such as aircraft parts, automobile parts, piping, welded components, and semiconductor devices need to be inspected during manufacturing and periodic inspections. In such cases, an inspection method is used in which part or the whole of the object to be inspected is photographed with an X-ray imaging device to check for defects, dimensions, and shape. During inspections, imaging is also performed periodically for the purpose of evaluating the imaging plate (IP) of the X-ray imaging device itself. For example, Patent Document 1 (Japanese Patent Application Publication No. 2024-19042) describes an X-ray diagnostic device comprising a setting unit for setting multiple regions of interest, a calculation unit for calculating statistical values ​​using the pixel values ​​of each region of interest, a threshold acquisition unit for acquiring a threshold based on the statistical values, and an X-ray condition determination unit for determining the X-ray conditions for imaging the next X-ray image based on the statistical values ​​and the threshold.

[0003] In the technology described in Patent Document 1 above, the user manually sets multiple regions of interest for the captured image, making it difficult to efficiently set the regions of interest.

[0004] This disclosure has been made in consideration of the above circumstances and aims to provide an image processing device, an image processing method, and an image processing program that can streamline the setting of regions of interest in captured images in X-ray inspections.

[0005] To achieve the above objective, an image processing apparatus in the first aspect of the present disclosure includes a processor, which receives the position of a target object in a captured image, sets candidate regions of interest from the captured image based on the position of the target object, calculates an image quality evaluation value which is an index value for evaluating the image quality of the captured image for each of the candidate regions of interest, and selects a region of interest from the candidate regions of interest of the captured image based on the image quality evaluation value.

[0006] In a second aspect of the present disclosure, the image processing apparatus, in the first aspect, further detects the position of the target object from the captured image.

[0007] In a third aspect of the present disclosure, the image processing apparatus, in the first aspect, includes a processor that sets candidate regions of interest based on the degree of change in pixel values ​​in a predetermined region of the captured image.

[0008] In a fourth aspect of the present disclosure, the image processing apparatus, in the first aspect, sets a region including the surface on which the object is placed as a candidate region of interest.

[0009] An image processing apparatus in a fifth aspect of the present disclosure, in an image processing apparatus in a first aspect, the processor changes at least one of the position and shape of the candidate region of interest and calculates the image quality evaluation value for at least one of the candidate region of interest before and after the change.

[0010] An image processing apparatus according to a sixth aspect of the present disclosure, in an image processing apparatus according to a fifth aspect, the processor selects a region of interest in the captured image based on the image quality evaluation value for at least one of the candidate region of interest before and after the change.

[0011] An image processing apparatus according to the seventh aspect of the present disclosure, in any one of the first to sixth aspects of the present disclosure, the processor calculates the image quality evaluation value of the candidate region of interest based on at least one of the image quality evaluation value of the candidate region of interest and the image quality evaluation value of the target object.

[0012] An image processing apparatus according to the eighth aspect of the present disclosure, in an image processing apparatus according to the seventh aspect, the processor calculates an image quality evaluation value of the target object based on a region image which is an image of the region including the target object.

[0013] An image processing apparatus according to the ninth aspect of this disclosure, in an image processing apparatus according to the seventh aspect, the processor presents the image quality evaluation values ​​obtained in a time series.

[0014] In the image processing apparatus of the tenth aspect of the present disclosure, in the image processing apparatus of the first aspect, the processor draws and presents the region of interest selected from the candidate regions of interest on the captured image.

[0015] An image processing apparatus according to the eleventh aspect of the present disclosure, in the image processing apparatus according to the tenth aspect, presents the image quality evaluation value in association with the region of interest selected from the candidate region of interest.

[0016] An image processing method according to a twelfth aspect of this disclosure involves a computer receiving the position of a target object in a captured image, setting candidate regions of interest from the captured image based on the position of the target object, calculating an image quality evaluation value which is an index value for evaluating the image quality of the captured image for each of the candidate regions of interest, and selecting a region of interest from the candidate regions of interest based on the image quality evaluation value.

[0017] The image processing program in the 13th aspect of this disclosure receives the position of a target object in a captured image, sets candidate regions of interest from the captured image based on the position of the target object, calculates an image quality evaluation value which is an index value for evaluating the image quality of the captured image for each of the candidate regions of interest, and causes the computer to perform the process of selecting a region of interest from the candidate regions of interest of the captured image based on the image quality evaluation value.

[0018] According to this disclosure, it is possible to streamline the setting of regions of interest in images during X-ray examinations.

[0019] This is a block diagram showing an example of the overall configuration of a radiation imaging system according to an embodiment. This is a block diagram showing an example of the configuration of the main part of the control device. This is a functional block diagram showing the functional configuration of the CPU in the control device of the radiation imaging system according to an embodiment. This is a diagram showing an example of an X-ray image in which multiple ROI candidates have been set. This is a perspective view showing an example of a Duplex Plate Phantom that is the target of imaging. This is a diagram showing an example of region extraction and region detection of Duplex IQI and Hole type IQI. This is a diagram showing another example of region extraction of Duplex IQI and Hole type IQI. This is a diagram showing another example of region detection of Duplex IQI and Hole type IQI. This is a flowchart showing an example of the processing flow by a program according to an embodiment. This is a conceptual diagram of a bandpass filter used for frequency analysis. This is a diagram for explaining the method of setting ROI candidates. This figure shows an example of a solid region detected from an X-ray image using machine learning. This figure shows an example of a solid region detected from an X-ray image using machine learning. This figure illustrates how to set ROI candidates on the Duplex Plate Phantom Plate. This figure shows an example of how the IQI is placed on the weld base material. This figure illustrates how to search for perturbation ROI candidates. This figure shows an example of the Duplex IQI profile. This figure shows the dip value and iSR of the Duplex IQI. ｂ This figure illustrates the calculation method. This figure shows an example of calculating SNR using Duplex Plate Phantom Type 2. This figure shows an example of image quality evaluation values ​​calculated from Duplex Plate Phantom Type 1. This is a graph showing an example of the time-series change of SNR. This figure shows an example of a method for presenting the selected ROI. This figure shows an example of an X-ray image in which multiple evaluation value information is superimposed. This figure shows an example of the flow of the image processing method according to the embodiment.

[0020] The embodiments of this disclosure will be described in detail below with reference to the drawings. However, these embodiments are not intended to limit this disclosure.

[0021] In this embodiment, a radiation image imaging system including a control device as an example of an image processing device will be described as an example. FIG. 1 is a block diagram showing an example of the overall configuration of the radiation image imaging system according to this embodiment.

[0022] As shown in FIG. 1, the radiation image imaging system 10 includes a radiation irradiation device 12, a radiation image imaging device 16, and a control device 18.

[0023] The radiation irradiation device 12 according to this embodiment includes a radiation source 14 that irradiates a subject W, which is an example of an imaging target, with radiation R such as X-rays. Note that the method of instructing the irradiation of radiation R to the radiation irradiation device 12 is not particularly limited. For example, when the radiation irradiation device 12 includes an irradiation button or the like, a user such as a radiologic technologist may irradiate the radiation R from the radiation irradiation device 12 by instructing the irradiation of the radiation R using the irradiation button. Also, for example, a user such as a radiologic technologist may irradiate the radiation R from the radiation irradiation device 12 by operating the control device 18 to instruct the irradiation of the radiation R.

[0024] When the radiation irradiation device 12 receives an instruction to irradiate the radiation R, it irradiates the radiation R from the radiation source 14 according to irradiation conditions such as the set tube voltage, tube current, and irradiation time. Hereinafter, the dose of the radiation R will be referred to as the "radiation dose".

[0025] The radiation image imaging device 16 according to this embodiment includes a radiation detector 20 (also referred to as an imaging plate) that detects the radiation R irradiated from the radiation irradiation device 12 and transmitted through the subject W. The radiation image imaging device 16 uses the radiation detector 20 to capture a radiation image of the subject W.

[0026] As shown in FIG. 2, the control device 18 includes a control unit 22, a display unit 32, an operation unit 34, and a communication I / F (interface) unit 35. FIG. 2 is a block diagram showing an example of the configuration of the main part of the control device 18.

[0027] The main part of the control unit 22 has a general computer configuration and controls the overall operation of the control device 18.

[0028] The control unit 22 includes a CPU 24, a ROM 26, and a RAM 28. The ROM 26 stores in advance various programs 30 and the like that are executed by the CPU 24, such as a shooting processing program and a display processing program executed during shooting control. The RAM 28 temporarily stores various data.

[0029] The display unit 32 displays a screen for operating the radiation irradiation device 12 and the radiation imaging device 16, and the captured radiation image and the like. For example, a direct-view type electronic display is applied to the display unit 32.

[0030] The operation unit 34 is used for the user to input instructions regarding shooting and various information. The operation unit 34 is not particularly limited, and examples include various switches, touch panels, touch pens, and mice.

[0031] The communication I / F unit 35 communicates captured images such as radiation images and various information with the radiation irradiation device 12 and the radiation imaging device 16 by wireless communication or wired communication.

[0032] By the way, during radiation imaging for medical and non-destructive inspection purposes, a chart is imaged to perform appropriate diagnosis and inspection, the image is analyzed to obtain resolution, contrast, etc., and the captured radiation image is evaluated. Also, based on the evaluation result, the radiation image is corrected and the radiation imaging device 16 is adjusted. For example, during radiation imaging for non-destructive inspection, an IQI (Image Quality Indicator) is used as a chart for evaluating the image. The types of IQI include Duplex type (duplex image quality meter), Hole type (perforated penetrometer), Wire type (wire penetrometer), etc.

[0033] In this case, before starting the examination, imaging may be performed to confirm whether the radiation detector 20 is functioning correctly. For example, instead of the subject W, the IQI is placed on a Duplex Plate Phantom and imaging is performed. In this case, since the region of interest (hereinafter referred to as "ROI (Region Of Interest)") is set manually on the same plane as the IQI, it is difficult to set the ROI efficiently.

[0034] Therefore, in this embodiment, the control device 18 automatically sets the ROI for the X-ray image based on the position of the IQI in the X-ray image. Note that the X-ray image is an example of an acquired image, and the IQI is an example of a target object (also called a "test object"). Specifically, the CPU 24 executes the program 30 stored in the ROM 26 of the control unit 22 by writing it to the RAM 28, thereby executing the functions shown in Figure 3. Figure 3 is a functional block diagram showing the functional configuration of the CPU 24 in the control device 18 of the radiation imaging system 10 according to this embodiment.

[0035] In other words, as shown in Figure 3, the CPU 24 functions as a reception unit 24A, a setting unit 24B, a calculation unit 24C, a selection unit 24D, and a presentation unit 24E.

[0036] The reception unit 24A receives input of an X-ray image and the location of IQIs in the X-ray image. Specifically, the reception unit 24A receives input of IQI locations that are automatically detected by image processing of the X-ray image, for example. Alternatively, the reception unit 24A may also accept IQI locations through manual operation by the user.

[0037] The setting unit 24B sets ROI candidates from the X-ray image based on the position of the IQI received by the receiving unit 24A. Specifically, the setting unit 24B sets ROI candidates based on the degree of change in pixel values ​​in a predetermined area of ​​the X-ray image. For example, indicator values ​​such as change amount and change rate can be used to represent the degree of change in pixel values. As ROI candidates, for example, areas without steep changes, that is, areas with a relatively small degree of change in pixel values ​​(so-called solid areas) are set. Various methods such as frequency analysis, edge filtering, MSER (Maximally Stable Extreme Regions), and machine learning can be used to set ROI candidates. The setting unit 24B may also set an area including the surface on which the IQI is placed as an ROI candidate. The surface on which the IQI is placed is, for example, a solid area with uniform pixel values.

[0038] The calculation unit 24C calculates an image quality evaluation value, which is an index value for evaluating the image quality of an X-ray image, for each of the ROI candidates (or each of the multiple ROI candidates if there are multiple ROI candidates) set by the setting unit 24B. Specifically, the calculation unit 24C changes at least one of the position and shape of the ROI candidate and calculates an image quality evaluation value for at least one of the ROI candidate before and after the change. Alternatively, the calculation unit 24C may calculate the image quality evaluation value of the ROI candidate based on at least one of the image quality evaluation value of the ROI candidate and the image quality evaluation value of the IQI. In this case, the calculation unit 24C calculates the image quality evaluation value of the IQI based on a region image, which is an image of the region containing the IQI.

[0039] The selection unit 24D selects an ROI from among the ROI candidates based on the image quality evaluation value calculated by the calculation unit 24C. Specifically, the selection unit 24D selects an ROI from the X-ray image based, for example, on the image quality evaluation value for at least one of the ROI candidates before and after the change.

[0040] The display unit 24E visualizes the image quality evaluation values ​​obtained in a time series and displays them on the display unit 32. Alternatively, the display unit 24E may also display a selected ROI from among the ROI candidates by drawing it on the X-ray image. In this case, the display unit 24E may also display the image quality evaluation values ​​associated with the selected ROI from among the ROI candidates.

[0041] Figure 4 shows an example of an X-ray image with multiple ROI candidates 1-4 configured. The X-ray image shown in Figure 4 is an image of a Duplex Plate Phantom Type 1 used for imaging plate (IP) evaluation. The X-ray image shows three IQIs (two Hole-type IQIs and a Duplex IQI) on two Plates 1 and 2, rather than the object being inspected, such as a metal part. In addition, four ROI candidates 1-4 are configured on Plates 1 and 2 as an example.

[0042] Here, we will explain the relationship between ROI and image quality evaluation. Duplex IQI image quality evaluation values ​​include, for example, dip value (an index value representing modulation relative to the magnitude of two peaks in an image), iSR ｂ (Interpolated basic spatial resolution) is used. Hole type IQI image quality evaluation values ​​include, for example, CNR (Contrast to Noise Ratio) and CS ４Ｔ (Contrast Sensitivity in 4T hole) is used. For ROI, for example, SNR (Signal to Noise Ratio) is used. Also, for the combination of Duplex IQI and ROI, for example, SNR Ｎ The normalized signal-to-noise ratio is used. Since these image quality evaluation values ​​can be calculated using known methods, a detailed explanation of the calculation method will be omitted.

[0043] Figure 5 is a perspective view showing an example of a Duplex Plate Phantom, which is the object of imaging. Plates 1 and 2 of the Duplex Plate Phantom shown in Figure 5 are provided with one Duplex IQI and two Hole-type IQIs. Plate 2 is thicker than Plate 1. When the Duplex Plate Phantom shown in Figure 5 is imaged, an X-ray image like the one shown in Figure 4 is obtained. However, the positions of the ROI candidates and IQIs are different between the Duplex Plate Phantom shown in Figure 5 and the X-ray image shown in Figure 4.

[0044] The reception unit 24A automatically detects the locations of two Hole-type IQIs from the X-ray image shown in Figure 4, for example. Known techniques are used to detect the locations of Hole-type IQIs, such as machine learning, graph cuts, and template matching.

[0045] Figure 6 shows an example of region extraction and region detection for Duplex IQI and Hole-type IQI. The left side of Figure 6 shows an example of region extraction for Duplex IQI and Hole-type IQI, and the right side of Figure 6 shows an example of region detection for Duplex IQI and Hole-type IQI. In the region extraction shown in the left side of Figure 6, the positions of Duplex IQI and Hole-type IQI can be extracted on a pixel-by-pixel basis. On the other hand, in the region detection shown in the right side of Figure 6, the positions of the four endpoints and center point of the rectangle (shown by a dotted line) surrounding the Duplex IQI can be detected, and the positions of the four endpoints and center point of the rectangle (shown by a solid line) surrounding the Hole-type IQI can be detected.

[0046] Figures 7 and 8 show another example of region extraction for Duplex IQI and Hole-type IQI. In the left diagram of Figure 6 above, all regions of Duplex IQI and Hole-type IQI are extracted, but as shown in Figures 7 and 8, only a portion of the regions of Duplex IQI and Hole-type IQI may be extracted. In the example in Figure 7, the portion of Hole-type IQI indicated by hatching is extracted. In the example in Figure 8, the portion of Duplex IQI indicated by hatching is extracted.

[0047] Figure 9 shows another example of region detection for Duplex IQI and Hole-type IQI. In the right diagram of Figure 6 above, the entire region of the Duplex IQI and Hole-type IQI is detected, but as shown in Figure 9, only a portion of the region of the Duplex IQI and Hole-type IQI may be detected. In the region detection in Figure 9, the positions of the four endpoints and center point of the rectangle (shown by the dotted line) enclosing a portion of the Duplex IQI are detected, and the positions of the four endpoints and center point of the rectangle (shown by the solid line) enclosing a portion of the Hole-type IQI are detected.

[0048] Next, with reference to Figure 10, the operation of the control device 18 according to this embodiment will be described.

[0049] Figure 10 is a flowchart showing an example of the processing flow by the program 30 according to this embodiment.

[0050] First, when the control device 18 receives an instruction to start the image quality evaluation, the CPU 24 reads and executes the program 30.

[0051] In step S101 of Figure 10, the CPU 24 receives input of an X-ray image and the location of the IQI in the X-ray image. Specifically, as an example, the CPU 24 automatically detects the location of the IQI by image processing of the X-ray image, as shown in Figures 6 to 9 above.

[0052] In step S102, the CPU 24 sets ROI candidates from the X-ray image based on the IQI positions received in step S101, as shown in Figure 4 above, as an example. Specifically, various methods such as frequency analysis, edge enhancement filters, MSER, and machine learning are used to set ROI candidates.

[0053] Here, ROI candidates are defined as so-called solid regions, that is, "regions where there are no abrupt changes in pixel values ​​(image patterns)." Figures 11 to 14 will be used to explain typical methods for defining ROI candidates.

[0054] Figure 11 is a conceptual diagram of a bandpass filter used for frequency analysis. The solid region, which is a candidate for ROI, is a region where the brightness changes smoothly and the spatial frequency is low. For this reason, the desired ROI candidate may be set by frequency filtering using the Fourier transform of the X-ray image. Specifically, methods such as excluding areas with high-frequency components or using a bandpass filter to allow only the range of intermediate frequency components to pass through can be considered. For example, as shown in Figure 11, by applying a bandpass filter to the X-ray image, low-frequency X-ray irradiation unevenness (A1), large (A2) and small (A3) medium-frequency stains, and high-frequency metal / IP material textures (A4) can be removed. In this case, the ROI candidate for the solid region is set from the medium-frequency X-ray image after passing through the bandpass filter.

[0055] Figure 12 illustrates the method for setting ROI candidates. When setting ROI candidates, ensure that they do not overlap with IQI, that is, that they are in an area outside the IQI region. Also, ensure that they do not include metal parts, pipe joints, point structures, or line structures. For example, as shown in Figure 12, if an ROI candidate includes a point structure, line structure, or joint, it is desirable to shift the ROI candidate to a nearby area, avoiding these structures. To check whether an ROI candidate is a solid region, for example, one method is to examine the luminance value distribution of the ROI candidate and extract the upper and lower X% {X | (0, 50), where X is a real number} regions. If the difference between the area of ​​the extracted region and the area of ​​the ROI candidate is smaller than a certain threshold, it can be determined that some kind of structure is included. Hereinafter, this method, that is, the method for checking whether or not an area is a solid region, will be referred to as the "solid region confirmation method".

[0056] Alternatively, an edge enhancement filter may be used. Since solid regions have no edges, an edge enhancement filter may be used to consider regions without edges as solid regions and set ROI candidates accordingly. For example, known filters such as the Sobel filter, the Canny algorithm (Gaussian filter + Sobel filter), the Laplacian filter, and the Hessian filter can be used as edge enhancement filters. With this method, ROI candidates can be set while avoiding the detected IQI and edges, and the solid region verification method described in Figure 12 above can be performed.

[0057] Alternatively, MSER may be used. MSER is a method for detecting stable regions of an image (e.g., an X-ray image) using a blob (binary large object) detector. Solid regions may be detected as MSER regions, or solid regions may be detected from regions where MSER regions are not densely concentrated. When setting ROI candidates from among solid regions, the solid region confirmation method described in Figure 12 above may be combined.

[0058] Alternatively, machine learning can be used. In this case, the raw data regions can be pre-trained using machine learning, allowing for automatic detection of these regions. Instance segmentation can also be used to distinguish objects within each raw data region.

[0059] Figures 13 and 14 show examples of solid regions detected from X-ray images using machine learning. The left figure of Figure 13 shows an example of detecting a Plate region where IQIs are placed, the right figure of Figure 13 shows an example of simultaneously detecting an Imaging Plate region, and Figure 14 shows an example of detecting all solid regions in the X-ray image. In Figures 13 and 14, the hatched areas indicate the detected regions.

[0060] Alternatively, the lowest Imaging Plate region can be extracted with a separate label when viewing the image in the depth direction, and the solid areas can be layered. Here, "layering" means that Plate 1 and 2 are contained within the Imaging Plate. When layered, it is possible to determine which plate is thicker by comparing the brightness values ​​of the Imaging Plate and Plate 1 and 2.

[0061] Alternatively, as shown in Figure 14, all solid regions in the X-ray image may be detected with separate labels, and ROI candidates may be set using the smallest solid region that contains a Hole-type IQI.

[0062] Here, we assume that the IQI is placed on various objects. That is, the IQI may be placed directly on the Imaging Plate, or it may be placed on, for example, the Phantom Plate, the base material for welding, or the planar area of ​​a metal part. Referring to Figure 15, we will explain an example where the IQI is placed on the Duplex Plate Phantom Plate.

[0063] Figure 15 illustrates a method for setting ROI candidates on the plates of Duplex Plate Phantom. The example in Figure 15 conceptually shows how to set ROI candidates on Plates 1 and 2 based on the detected Hole Type IQI and Plates 1 and 2.

[0064] As shown in (B1) of the left diagram of Figure 15, ROI candidates may be set based on predetermined coordinates of Plate 1 (for example, the coordinates of the upper left endpoint), or ROI candidates may be set based on predetermined coordinates of the Hole-type IQI of Plate 1 (for example, the coordinates of the upper left endpoint). The coordinates of the ROI candidates are, for example, (+x, +y) when the upper left endpoint of Plate 1 is used as the reference, and (-x, -y) when the upper left endpoint of the Hole-type IQI of Plate 1 is used as the reference. Alternatively, as shown in (B2) of the left diagram of Figure 15, ROI candidates may be set above and below the predetermined coordinates of the Hole-type IQI of Plate 1 (for example, the center coordinates). Furthermore, as shown in (B3) of the right-hand figure of Figure 15, ROI candidates may be set at predetermined points (e.g., the midpoint) on the line segment connecting the characteristic coordinates of Hole Type IQI and Plate1 (e.g., the coordinates of the respective endpoints, the coordinates of the midpoint, etc.).

[0065] When two ROI candidates are set for each Plate, the two ROI candidates are often set at distant locations. As shown in (B4) of the left diagram of Figure 15, Plate 2 may be divided into N x N (for example, N=3) sections, and ROI candidates may be set at distant locations (for example, the diagonally opposite upper left and lower right). Alternatively, as shown in (B5) of the right diagram of Figure 15, Plate 2 may be divided into N (for example, N=4) sections based on a predetermined point of the Hole Type IQI (for example, the center point), and ROI candidates may be set at distant locations (for example, the diagonally opposite upper left and lower right). If an ROI candidate overlaps with an IQI (B5), it can be reset to a non-overlapping position within the region where the ROI candidate is located (for example, the lower right region).

[0066] Furthermore, MSER may be used to set ROI candidates. Various objects can be used as targets for IQI placement. However, since it is difficult to calculate image quality evaluation values ​​for the purpose of IQI placement on a plane where pixel values ​​change in a complex manner, it is assumed that the local plane on which the IQI is placed has uniform pixel values. Therefore, the MSER region may be detected, and the region containing the IQI may be used as the IQI placement surface. If the MSER region is not detected, the MSER parameters may be changed.

[0067] Furthermore, machine learning may be used to define ROI candidates. In this case, the machine learning is trained to recognize objects in which IQIs are placed, regardless of whether they are solid regions or not. For example, a SAM (Segment Anything Model) can be used to recognize all objects in an X-ray image and search for objects that contain IQIs. This machine learning method is also considered useful, for example, when grayscale occurs in the plate due to dose considerations.

[0068] Figure 16 shows an example of how an IQI is placed on a welding base material. In the example in Figure 16, a Hole-type IQI is placed on the welding base material. In the lower part of Figure 16, the base material on which the Hole-type IQI with the number "5" is placed is represented by hatching. In this hatched area, ROI candidates are set while avoiding IQIs and other objects. Here, as a reference example, the size of ROI candidates is defined in the welding standard ISO / EN 17636-2 as having 20 pixels on the short side and 55 pixels or more on the long side. ROI candidates can be either horizontal or vertical.

[0069] Furthermore, the installation surface of the IQI may be the plate surface of the Duplex Plate Phantom shown in Figure 4 above.

[0070] Furthermore, the installation surface of the IQI may be the Imaging Plate surface. If the IQI is placed directly on the Imaging Plate, the installation plane will be the entire image. Candidate ROIs should be set so as to avoid the IQI within the entire image. Basically, point-like and linear structures do not appear in the X-ray image of the Imaging Plate itself, but it is desirable to confirm whether or not it is a solid region.

[0071] Furthermore, the mounting surface of the IQI may be a flat area of ​​the component. For example, the IQI may be placed on a curved surface of a metal component such as a turbocharger in an automobile.

[0072] Next, with reference to Figure 17, a method for changing at least one of the position and shape of a candidate ROI will be specifically described. Hereinafter, changing at least one of the position and shape of a candidate ROI will be referred to as "perturbation of the candidate ROI," and the ROI candidate after perturbation will be referred to as the "perturbed candidate ROI."

[0073] Figure 17 illustrates a method for searching for perturbed ROI candidates. Multiple perturbed ROI candidates are searched by perturbing the ROI candidates, and the final ROI is selected after image quality evaluation and ROI candidate selection. The method for searching for perturbed ROI candidates is as shown in Figure 17. As shown in (C1) of Figure 17, perturbed ROI candidates may be searched by changing the position of ROI candidate 1. As shown in (C2) of Figure 17, perturbed ROI candidates may be searched by changing the size of ROI candidate 2. As shown in (C3) of Figure 17, perturbed ROI candidates may be searched by rotating ROI candidate 3. As shown in (C4) of Figure 17, perturbed ROI candidates may be searched by changing both the position and size of ROI candidate 4.

[0074] In other words, based on the position and size of each ROI candidate that has been set once, perturbed ROI candidates are searched for while changing at least one of the position, size, and rotation angle of each ROI candidate. Then, for each perturbed ROI candidate or ROI candidate, an example of an image quality evaluation value, the SNR is calculated, and the perturbed ROI candidate or ROI candidate with the maximum SNR may be selected as the final ROI.

[0075] Returning to Figure 10, in step S103, the CPU 24 calculates an image quality evaluation value, which is an index value for evaluating the image quality of the X-ray image for each of the ROI candidates (or each of the multiple ROI candidates if there are multiple ROI candidates) set in step S102. As mentioned above, for example, the image quality evaluation value for Duplex IQI includes, for example, dip value, iSR ｂ The following are used. Hole type IQI image quality evaluation values ​​include, for example, CNR and CS. ４Ｔ For example, SNR is used as the ROI. Also, for example, SNR is used in combinations of Duplex IQI and ROI. Ｎ This is used.

[0076] Figure 18 shows an example of a Duplex IQI profile. The iSR is calculated from the dip value using Duplex IQI. ｂIt can be calculated. The detection result of the Duplex IQI is obtained as the coordinates of the start point and the end point of the Duplex IQI. The start point and the end point of the Duplex IQI are determined as the start point and the end point of a line segment that passes through the center of the Duplex IQI in the longitudinal direction of the Duplex IQI. As shown in the upper figure of FIG. 18, an area to be used in the calculation is displayed on the Duplex IQI based on the obtained coordinates of the start point and the end point of the Duplex IQI. There are a plurality of line pairs between the start point and the end point of the Duplex IQI, and a pixel value profile is created and analyzed. This profile is not obtained from one line connecting the start point and the end point, but is created by averaging the data of a plurality of upper and lower lines.

[0077] FIG. 19 is a diagram for explaining a method of calculating the dip value and iSR of the Duplex IQI. ｂ The calculation method shown in FIG. 19 is, for example, a known method described in JIS Z 3110. In FIG. 19, (E1) shows an example of an image of a Duplex IQI (duplex image quality indicator) in a digital image, and (E2) shows an example of a profile of the Duplex IQI averaged from a plurality of lines (for example, 21 or more lines). (E3) shows an example of an enlarged profile of line pairs D7 and D8, and (E4) shows an example of a dip value. (E5) is a diagram showing an example of the correspondence between the dip value and SR ｂ is a diagram showing an example of the correspondence relationship with.

[0078] As shown in (E1) to (E4) of FIG. 19, a dip value is obtained from the profile of the Duplex IQI. That is, as shown in (E4) of FIG. 19, using the following formula, the dip value (%) is calculated using the lengths (A, B, C) of the valleys (that is, peaks) of each line pair.

[0079] dip = 100×(A + B - 2C) / (A + B)

[0080] At the actual inspection site, it is confirmed which is the first line pair when the dip value of the line pair first becomes Y1% or less. As the value of Y1, for example, 20% may be used. Next, as shown in (E5) of FIG. 19, a spatial resolution is associated with each line pair from the width between the line pairs, and an approximate formula for the spatial resolution and the dip value is created. And the iSR at the time of Y1%ｂ The value is determined by the intersection with the Y1% line, and in this example, iSR ｂ = X1 μm.

[0081] Next, we will explain how to determine the SNR from ROI candidates. First, we calculate the average pixel count of the ROI candidates and then the standard deviation of the ROI candidates. The average pixel count or standard deviation of the ROI candidates can be calculated for the entire ROI candidate, or for each row, or for each column. The standard deviation of the ROI candidates is calculated using the following formula, where S is the standard deviation and x is x. － (- is directly above x) is the mean, n is the number of data points, x ｉ The values ​​of each data point are shown.

[0082]

[0083]

[0084] Next, we calculate the SNR. Figure 20 shows an example of calculating the SNR using Duplex Plate Phantom Type 2. In the example in Figure 20, the following formula is used to calculate the SNR from ROI candidates 1 to 4. ｔｈｉｎ and SNR ｔｈｉｃｋ Calculate the following. However, PV ｍｅａｎ σ represents the mean pixel value, and Sigma represents the standard deviation, which is the square root of the variance. Simply put, if I is the mean of the pixel values ​​and σ is the standard deviation, then SNR = I / σ. ｔｈｉｎ This represents the SNR of Plate 1, which is the thinner plate. ｔｈｉｃｋ This represents the SNR of Plate 2, which has a greater thickness.

[0085]

[0086]

[0087] Next, SNR Ｎ This is calculated using the following formula: SNR obtained from ROI candidates 1-4 and iSR obtained from Duplex IQI. ｂ Use this. Note that SNR Ｎ If you only need to find the maximum and minimum values, use iSR. ｂ It's not necessary; the SNR alone is sufficient.

[0088] SNR Ｎ = SNR 測定値 ×88.6μm / iSR ｂ

[0089] Figure 21 shows the image quality evaluation values ​​obtained from each IQI and ROI candidate of Duplex Plate Phantom.

[0090] Figure 21 shows an example of image quality evaluation values ​​calculated from Duplex Plate Phantom Type 1. The SNR values ​​(8) to (11) shown in Figure 21 Ｎ In addition, for example, from (1), (4), and (5) SNR Ｎ You can also find the SNR from (1), (6), and (7) Ｎ You may also calculate (8) to (11). However, it is not necessary to calculate all of (8) to (11). If you do not calculate all of them, determine the best SNR using (4) to (7) and use the corresponding ROI candidate to find the largest SNR. Ｎ You just need to decide on that. Alternatively, you can select the ROI candidate with the highest SNR for each plate, and then select the ROI with the highest SNR for each plate. Ｎ It can also be calculated.

[0091] Returning to Figure 10, in step S104, the CPU 24 selects an ROI for the X-ray image from among the ROI candidates based on the image quality evaluation value calculated in step S103, and terminates the series of processes by this program 30. Specifically, the CPU 24 selects an ROI for the X-ray image based, for example, on the image quality evaluation value for at least one of the ROI candidates before and after the change.

[0092] Here, the image quality evaluation values ​​can be monitored over time using a graph. In this case, the image quality evaluation values ​​obtained sequentially are maintained and the progress is visualized and presented. For example, in the case of Duplex Plate Phantom, since it is performed daily for the purpose of evaluating Imaging Plate, the time-series changes can be graphed.

[0093] Figure 22 is a graph showing an example of the time-series change in SNR. In Figure 22, the horizontal axis represents the date and the vertical axis represents the SNR. The SNR on 2024 / 7 / 4 shows an abnormal value. In this case, it may be possible to use the rate of change over the past few days to trigger an alert notification when acquiring or evaluating X-ray images.

[0094] Figure 23 shows an example of a method for presenting selected ROIs. The selected ROI may be presented by drawing it on or next to the examination image. In the example in Figure 23, the case of presenting an ROI on an endoscopic image is shown as a reference example. As shown in (Presentation 1) of Figure 23, the ROI may be presented by surrounding it on the image. If it is difficult to see when overlapping, a method of placing a mask image of the detection result next to it may be used, as shown in (Presentation 2) of Figure 23.

[0095] Figure 24 shows an example of an X-ray image in which multiple evaluation value information is superimposed. In the example of Figure 24, multiple selected ROIs are presented. The multiple selected ROIs may be presented in pairs with the evaluation value information of the image quality evaluation value. Alternatively, a separate display list may be provided, allowing the user to select the image quality evaluation value to be superimposed on the image using checkboxes or the like. Furthermore, if the maximum and minimum (best and worst) values ​​of the image quality evaluation value are calculated, the user may be able to select which one to superimpose on the image. Here, it is desirable to exclude areas containing objects, linear structures, etc., from the display position of the evaluation value information and adjust the display position so that they do not overlap. Also, if the evaluation value information and ROI overlap, the display position of the evaluation value information may be reset.

[0096] Furthermore, when the solid region is small and the ROI overlaps with the IQI, the position and size may be automatically adjusted. Since a larger ROI size is desirable for calculating the SNR, the position of the ROI with the largest size may be automatically determined.

[0097] Figure 25 shows an example of the flow of the image processing method according to this embodiment. In Figure 25, rectangles represent data and circles represent processing.

[0098] In S1 of Figure 25, the CPU 24 detects the position of an IQI (Duplex IQI, Hole-type IQI), which is an example of a test object, from an X-ray image, which is an example of an captured image.

[0099] In S2, the CPU 24 calculates multiple ROI candidates based on the location of the detected IQI. Specifically, it avoids the IQI and calculates solid areas in the image as ROI candidates.

[0100] In S3, CPU24 calculates the IQI image quality evaluation value. As mentioned above, the Duplex IQI image quality evaluation value includes, for example, dip value, iSR ｂ The image quality evaluation values ​​for Hole Type IQI include, for example, CNR and CS. ４Ｔ This is used.

[0101] In S4, the CPU 24 searches for multiple perturbed ROI candidates by changing at least one of the position and shape of each of the multiple ROI candidates, as explained in Figure 17 above, as an example.

[0102] In S5, the CPU 24 calculates the image quality evaluation value for each of the multiple perturbation ROI candidates from the multiple perturbation ROI candidates and the X-ray image. As mentioned above, for example, SNR is used as the image quality evaluation value for the perturbation ROI candidates.

[0103] In S6, the CPU 24 selects an ROI from among several perturbation ROI candidates based on the calculated image quality evaluation value.

[0104] In S7, the CPU 24 calculates the image quality evaluation value of the selected ROI based on the IQI image quality evaluation value. As mentioned above, the ROI image quality evaluation value includes, for example, SNR Ｎ This is used.

[0105] Thus, according to this embodiment, the region of interest in an image during X-ray inspection can be set automatically, making the process of setting the region of interest more efficient.

[0106] Although one embodiment of the control device 18 has been described above using the embodiments, the disclosed embodiment of the control device 18 is merely an example, and the embodiment of the control device 18 is not limited to the scope described in the embodiments. Various modifications or improvements can be made to the embodiments without departing from the gist of this disclosure, and such modified or improved embodiments are also included in the technical scope of the disclosure.

[0107] In the above embodiment, as an example, a configuration in which the control processing of the control device 18 is implemented by software processing was described. However, the control processing of the control device 18 may also be performed by hardware. In this case, the processing speed can be increased compared to the case in which it is implemented by software processing.

[0108] In the embodiments described above, the term "processor" refers to a broad type of processor, including general-purpose processors (e.g., CPUs) and specialized processors (e.g., GPUs: Graphics Processing Units, ASICs: Application Specific Integrated Circuits, FPGAs: Field Programmable Gate Arrays, programmable logic devices, etc.).

[0109] Furthermore, the operation of the processor in the above embodiment may not be performed by a single processor, but may be performed by multiple processors located in physically separate locations working together. Also, the order of the processor's operations is not limited to the order described in the above embodiment, but may be changed as appropriate.

[0110] In the above embodiment, an example was described in which the program 30 is stored in ROM 26. However, the storage location of the program 30 is not limited to ROM 26. The program 30 of this disclosure can also be provided in a form stored on a computer-readable storage medium. Alternatively, it may be provided in the form of a computer program product including the program 30. This disclosure can also be applied to programs and program products.

[0111] For example, program 30 may be provided in the form of a CD-ROM, DVD-ROM, or Blu-ray disc stored on an optical disc. Alternatively, program 30 may be provided in the form of a USB (Universal Serial Bus) memory or memory card stored on a portable semiconductor memory. These CD-ROMs, DVD-ROMs, Blu-ray discs, USBs, and memory cards are examples of non-transitor storage media.

[0112] The technology disclosed herein extends to all program products. A program product includes all forms of products for providing programs. For example, a program product includes programs provided via a network such as the Internet, and non-temporary computer-readable recording media such as CD-ROMs, DVDs, and USB memory sticks on which programs are stored.

[0113] The following additional information is disclosed regarding the embodiments described above.

[0114] (Note 1) An image processing device comprising a processor, wherein the processor receives the position of a target object in a captured image, sets candidate regions of interest from the captured image based on the position of the target object, calculates an image quality evaluation value which is an index value for evaluating the image quality of the captured image for the candidate regions of interest, and selects a region of interest from the candidate regions of interest of the captured image based on the image quality evaluation value. (Note 2) The image processing device according to Note 1, wherein the processor further detects the position of the target object from the captured image. (Note 3) The image processing device according to Note 1 or Note 2, wherein the processor sets candidate regions of interest based on the degree of change in pixel values ​​in a predetermined region of the captured image. (Note 4) The image processing device according to any one of Notes 1 to 3, wherein the processor sets a region including the surface on which the target object is placed as a candidate region of interest. (Note 5) The image processing apparatus according to Note 1, wherein the processor changes at least one of the position and shape of the candidate region of interest and calculates the image quality evaluation value for at least one of the candidate region of interest before and after the change. (Note 6) The image processing apparatus according to Note 5, wherein the processor selects the region of interest of the captured image based on the image quality evaluation value for at least one of the candidate region of interest before and after the change. (Note 7) The image processing apparatus according to any one of Notes 1 to 6, wherein the processor calculates the image quality evaluation value of the candidate region of interest based on at least one of the image quality evaluation value of the candidate region of interest and the image quality evaluation value of the target object. (Note 8) The image processing apparatus according to Note 7, wherein the processor calculates the image quality evaluation value of the target object based on a region image which is an image of the region including the target object. (Note 9) The image processing apparatus according to Note 7, wherein the processor presents the image quality evaluation values ​​obtained in time series. (Note 10) The image processing apparatus according to any one of Notes 1 to 9, wherein the processor displays the region of interest selected from the candidate regions of interest on the captured image.(Note 11) The image processing apparatus according to Note 10, wherein the processor presents the selected region of interest from the candidate regions of interest in association with the image quality evaluation value. (Note 12) An image processing method in which a computer performs the following processes: receiving the position of a target object in a captured image; setting candidate regions of interest from the captured image based on the position of the target object; calculating an image quality evaluation value, which is an index value for evaluating the image quality of the captured image, for the candidate regions of interest; and selecting a region of interest from the candidate regions of interest based on the image quality evaluation value. (Note 13) An image processing program for causing a computer to perform the following processes: receiving the position of a target object in a captured image; setting candidate regions of interest from the captured image based on the position of the target object; calculating an image quality evaluation value, which is an index value for evaluating the image quality of the captured image, for the candidate regions of interest; and selecting a region of interest from the candidate regions of interest based on the image quality evaluation value.

[0115] The disclosure of Japanese Patent Application No. 2024-160119, filed on 17 September 2024, is incorporated herein by reference in its entirety. All documents, patent applications, and technical standards described herein are incorporated herein by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

Claims

1. An image processing device comprising a processor, the processor receiving the position of a target object in a captured image, setting candidate regions of interest from the captured image based on the position of the target object, calculating an image quality evaluation value which is an index value for evaluating the image quality of the captured image for the candidate regions of interest, and selecting a region of interest from the candidate regions of interest of the captured image based on the image quality evaluation value.

2. The image processing apparatus according to claim 1, wherein the processor further detects the position of the target object from the captured image.

3. The image processing apparatus according to claim 1, wherein the processor sets the candidate region of interest based on the degree of change in pixel values ​​in a predetermined region of the captured image.

4. The image processing apparatus according to claim 1, wherein the processor sets a region including the surface on which the object in question is placed as the candidate region of interest.

5. The image processing apparatus according to claim 1, wherein the processor changes at least one of the position and shape of the candidate region of interest and calculates the image quality evaluation value for at least one of the candidate region of interest before and after the change.

6. The image processing apparatus according to claim 5, wherein the processor selects a region of interest in the captured image based on the image quality evaluation value for at least one of the candidate region of interest before and after the change.

7. The image processing apparatus according to any one of claims 1 to 6, wherein the processor calculates the image quality evaluation value of the candidate region of interest based on at least one of the image quality evaluation value of the candidate region of interest and the image quality evaluation value of the target object.

8. The image processing apparatus according to claim 7, wherein the processor calculates an image quality evaluation value of the target object based on a region image which is an image of the region including the target object.

9. The image processing apparatus according to claim 7, wherein the processor presents the image quality evaluation values ​​obtained in a time series.

10. The image processing apparatus according to claim 1, wherein the processor displays the region of interest selected from the candidate regions of interest on the captured image.

11. The image processing apparatus according to claim 10, wherein the processor presents the image quality evaluation value in association with the region of interest selected from the candidate regions of interest.

12. An image processing method in which a computer performs the following steps: receiving the position of a target object in a captured image; setting candidate regions of interest from the captured image based on the position of the target object; calculating an image quality evaluation value, which is an index value for evaluating the image quality of the captured image, for the candidate regions of interest; and selecting a region of interest from the candidate regions of interest of the captured image based on the image quality evaluation value.

13. An image processing program that causes a computer to perform the following processes: receive the position of a target object in a captured image; set candidate regions of interest from the captured image based on the position of the target object; calculate an image quality evaluation value, which is an index value for evaluating the image quality of the captured image for each of the candidate regions of interest; and select a region of interest from the candidate regions of interest of the captured image based on the image quality evaluation value.

Images