Alignment method, alignment device, program, and storage medium

Abstract

To provide an alignment method, an alignment device, a program, and a storage medium capable of more easily aligning an image of a design layout and an image obtained by photographing.SOLUTION: An alignment method according to an embodiment causes a computer to acquire a first preparatory image showing a design layout of a structure and a second preparatory image obtained by imaging the structure. A first parameter to be used in the first processing and a second parameter to be used in the second processing are determined based on the similarity between the first preparation image subjected to the first processing and the second preparation image subjected to the second processing. A first processed image is acquired by performing first processing on a first image showing a design layout using a first parameter. A second processed image is acquired by performing second processing on a second image obtained by photographing, using a second parameter. The first processed image and the second processed image are aligned based on the similarity between the first processed image and the second processed image.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to an alignment method, an alignment apparatus, a program, and a storage medium.

Background Art

[0002] There are cases where a photograph of a structure is taken and the photograph is aligned with an image of a design layout. There is a need for a technique that can more easily align two images.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The problem to be solved by the embodiments of the present invention is to provide an alignment method, an alignment apparatus, a program, and a storage medium that can more easily align an image of a design layout and an image obtained by photographing.

Means for Solving the Problems

[0005] The alignment method according to the embodiment causes a computer to acquire a first preparation image showing the design layout of a structure and a second preparation image obtained by photographing the structure. The alignment method causes the computer to determine a first parameter used in the first processing and a second parameter used in the second processing based on the similarity between the first preparation image that has undergone the first processing and the second preparation image that has undergone the second processing. The alignment method causes the computer to acquire a first processed image by applying the first processing to the first image showing the design layout of the structure using the first parameter. The alignment method causes the computer to acquire a second processed image by applying the second processing to the second image obtained by photographing the structure using the second parameter. The alignment method causes the computer to align the first processed image and the second processed image based on the similarity between the first processed image and the second processed image. [Brief explanation of the drawing]

[0006] [Figure 1] Figure 1 is a flowchart illustrating the alignment method according to the embodiment. [Figure 2] Figure 2 is a flowchart showing a part of the alignment method according to the embodiment. [Figure 3] Figure 3 is a flowchart showing a part of the alignment method according to the embodiment. [Figure 4] Figure 4 is a flowchart showing a part of the alignment method according to the first modified embodiment. [Figure 5] Figure 5 is a flowchart showing a part of the alignment method according to the first modified embodiment. [Figure 6] Figure 6 is a flowchart showing a part of the alignment method according to a second modified example of the embodiment. [Figure 7] Figure 7 is a flowchart showing a part of the alignment method according to a third modified example of the embodiment. [Figure 8]Figure 8 is a flowchart showing a part of the alignment method according to the fourth modified embodiment. [Figure 9] Figure 9 is a flowchart showing a part of the alignment method according to a second modified example of the embodiment. [Figure 10] Figure 10 is a schematic diagram illustrating the process shown in Figure 9. [Figure 11] Figure 11(a) is an example of the first partial image. Figure 11(b) is an example of the second partial image. Figure 11(c) is an example of the first preparation image. [Figure 12] Figure 12 shows an example of a second preparation image. [Figure 13] Figure 13 is a graph showing the relationship between the amount of resizing change and the number of image pairs. [Figure 14] Figure 14 is a graph showing the relationship between the rotation angle and the number of image pairs. [Figure 15] Figure 15 is a graph showing the relationship between the size of the smoothing filter and the similarity score. [Figure 16] Figure 16(a) is an example of a first-processed image after the first processing has been applied. Figure 16(b) is an example of a second-processed image after the second processing has been applied. [Figure 17] Figure 17 is an example of an image showing the test results. [Figure 18] Figure 18 is a schematic diagram representing the configuration of a computer. [Modes for carrying out the invention]

[0007] The embodiments of the present invention will be described below with reference to the drawings. In this specification and in the drawings, elements similar to those already described are denoted by the same reference numerals, and detailed descriptions will be omitted as appropriate.

[0008] When a structure is manufactured, it may be inspected. In the inspection, electrical characteristics, thermal characteristics, mechanical characteristics, etc. are examined. As an example, in the inspection of electrical characteristics, the electrical characteristics (such as electrical resistance, capacitance, etc.) of each point within a specific range of the structure are measured. In the inspection of thermal characteristics, the thermal characteristics (such as thermal conductivity, heat capacity, etc.) of each point within a specific range of the structure are measured. For example, as a result of the inspection, an image in which the characteristics of each point are mapped is obtained.

[0009] During the inspection, a photograph of the structure is also taken. The photograph is an optical image, an electronic image, an X-ray image, or an ultrasonic image. The shape of the structure does not appear sufficiently in the inspection image showing the inspection results. Therefore, it is difficult to determine where there is an abnormality in the structure only from the inspection image. The photograph of the structure can be used when identifying the abnormal location of the structure. By referring to the photograph, it becomes easier to identify the abnormal location.

[0010] However, when the structure is small, it may be difficult to identify which part of the structure has an abnormality only from the photograph. In particular, when the structure is a semiconductor device, each component is fine, and the structure of the structure does not clearly appear in the photograph. Depending on the manufacturing process, the manufactured shape may be different from the designed shape. Therefore, it may be difficult to identify the abnormal location only from the photograph.

[0011] When it is difficult to identify the abnormal location from the photograph, the design layout is further referred to. For example, the photograph is compared with the image of the design layout. The range shown in the photograph is identified as to which range of the design layout it corresponds to. Next, based on the result of aligning the photograph and the design layout, it is identified which part of the design layout the abnormal location in the inspection image corresponds to. Thereby, the abnormal location on the design layout is identified. The result of identifying the abnormal location is used for design or development to prevent the occurrence of the abnormality.

[0012] In recent years, semiconductor devices have become even more miniaturized. As a result, features such as shape and pattern tend to become less clear in the photographs taken. The less clear the features in the photograph, the more difficult it becomes to align the photograph with the design layout. Embodiments of the present invention are used to align an image showing a design layout with an image of a structure with higher precision.

[0013] Figure 1 is a flowchart illustrating the alignment method according to the embodiment. As shown in Figure 1, the alignment method according to the embodiment mainly includes parameter determination (step S10) and alignment (step S20). In step S10, the parameters necessary for alignment are determined. In step S20, the image of the design layout and the photograph of the structure are aligned. In step S20, the parameters determined in step S10 are used. Steps S10 and S20 are performed by a computer.

[0014] Figure 2 is a flowchart showing a part of the alignment method according to the embodiment. Figure 2 shows the details of step S10 in Figure 1. First, the computer acquires a first preparation image showing the design layout of the structure (step S11a). The first preparation image is the design layout converted into image data. For example, the first preparation image is prepared by converting two-dimensional Computer Aided Design (CAD) data or three-dimensional CAD data into an image. The computer may also acquire the first preparation image by acquiring CAD data and converting it into an image. Alternatively, the first preparation image may be prepared in advance, and the computer may acquire the prepared first preparation image.

[0015] The computer acquires a second preparation image obtained by photographing the structure (step S11b). The first and second preparation images relate to the same structure. For example, a camera photographs the structure and acquires a photograph (image) of the structure. The photograph may be an RGB color image or a grayscale image. For example, the camera receives infrared light and generates a grayscale image based on infrared light. The computer acquires an optical image of the structure as the second preparation image. If the structure is minute, the camera is preferably mounted on an optical microscope. Instead of an optical image, an electron image may be acquired. For example, an electron image of the structure may be acquired by a scanning electron microscope (SEM). In that case, the computer acquires the electron image as the second preparation image.

[0016] The extent of the structure shown in the first preparation image is approximately the same as the extent of the structure shown in the second preparation image. The extent shown in the design layout image may be slightly larger than the extent included in the photograph of the structure. It is preferable that the entire extent included in the photograph is included in the design layout image.

[0017] If the resolution of the first preparation image and the second preparation image are different, it is preferable to pre-adjust the resolution of at least one of the first and second preparation images. Generally, images of design layouts depict the shape of structures in more detail than photographs. That is, the resolution of design layout images is higher than the resolution of photographs. For this reason, it is preferable to pre-reduce the resolution of the first preparation image to reduce the difference between the resolution of the first and second preparation images.

[0018] The order of steps S11a and S11b can be changed as appropriate. Step S11b may be executed before step S11a, or steps S11a and S11b may be executed simultaneously.

[0019] The computer sets a first parameter used in the first process and a second parameter used in the second process (step S11c). The first process is image processing applied to an image of the design layout. The second process is image processing applied to an image of the structure.

[0020] For example, at least one of the first and second processes includes a smoothing process. In this case, at least one of the first and second parameters is the size of the smoothing filter. In the smoothing process, pixel values ​​are smoothed between adjacent pixels. The "smoothing filter" indicates the range over which pixel values ​​are smoothed. The larger the size of the smoothing filter, the wider the range over which pixel values ​​are smoothed. An averaging filter or a Gaussian filter can be used as the smoothing filter.

[0021] At least one of the first and second processes may include resizing (enlarging or reducing) the image. In this case, at least one of the first and second parameters is the amount of change in resizing. At least one of the first and second processes may include rotating the image. In this case, at least one of the first and second parameters is the angle. The first process may include multiple image processes. In this case, the first parameter is set for each image process. The second process may include multiple image processes. In this case, the second parameter is set for each image process.

[0022] In step S11c, the first and second parameters are set to their initial values. These initial values ​​may be pre-set by the computer user or randomly set by the computer.

[0023] The computer performs a first process on the first prepared image using the first parameter and a second process on the second prepared image using the second parameter (step S11d). The computer calculates the similarity between the first prepared image that has undergone the first process and the second prepared image that has undergone the second process (step S11e).

[0024] As a specific example, template matching is used to calculate similarity. The processed second preparation image is used as the template image. In the processed first preparation image, a range is set for calculating similarity with the template image. The similarity between the range set in the first preparation image and the second preparation image is calculated. The computer repeats the similarity calculation while changing the range for calculating similarity in the first preparation image. This ensures that similarity is calculated for the entire area of ​​the processed first preparation image. The computer extracts the highest similarity from the multiple similarities. The highest similarity is stored, linked to the first and second parameters at the time step S11e was performed. Feature point matching may be used instead of template matching for calculating similarity.

[0025] The computer determines whether the entire range of the first parameter and the entire range of the second parameter have been explored by that point (step S11f). For example, the range to be explored is predetermined for each of the first and second parameters. The computer determines whether all of the predetermined ranges have been explored.

[0026] If it is determined that the entire range has not been explored, the computer changes either the first or second parameter (step S11g). At this time, either only one of the first or second parameter may be changed, or both the first and second parameters may be changed.

[0027] After step S11g, step S11d is executed again. This allows the first and second processes to be performed using the modified first and second parameters, and the similarity is calculated. Steps S11d to S11g are repeated until the entire range of the first parameter and the entire range of the second parameter have been explored.

[0028] Instead of searching the entire range, the loop execution count may be set as the termination condition for steps S11d to S11g. In this case, the computer terminates the loop when the number of executions reaches a predetermined value. A threshold for similarity may also be set as a termination condition. In this case, the computer compares the similarity obtained in step S11e with a pre-set threshold. If the similarity exceeds the threshold, the computer terminates the loop. The loop is repeated until the similarity exceeds the threshold or the number of executions reaches a predetermined value.

[0029] Furthermore, in steps S11f and S11g, optimization methods other than exhaustive search may be used. For example, in the initial stages of the search, the similarity is calculated while significantly changing at least one of the first and second parameters. From the results, the computer calculates an approximation formula representing the relationship between the first parameter and the similarity, and an approximation formula representing the relationship between the second parameter and the similarity. From these approximation formulas, the computer determines the range in which the optimal first parameter is estimated to be obtained, and the range in which the optimal second parameter is estimated to be obtained. Within these ranges, the computer calculates the similarity while small changes are made to at least one of the first and second parameters.

[0030] If it is determined in step S11f that the entire range has been explored, the computer refers to the multiple similarity values ​​obtained. The computer extracts the highest similarity value from the multiple similarity values. The computer refers to the first and second parameters that were used when the highest similarity value was obtained. The computer determines these first and second parameters as the parameters to be used for the first and second processing (step S11h).

[0031] According to steps S11d to S11h, the first and second parameters are determined based on the similarity between the first prepared image, which has undergone the first processing, and the second prepared image, which has undergone the second processing. More specifically, the first and second parameters are determined such that the first prepared image, which has undergone the first processing, and the second prepared image, which has undergone the second processing, are similar.

[0032] Figure 3 is a flowchart showing a part of the alignment method according to the embodiment. Figure 3 shows details of step S20 in Figure 1. First, the computer acquires a first image showing the design layout of the structure (step S21a). The first preparation image and the first image relate to the same structure. The scope of the design layout in the first image may be the same as or different from the scope of the design layout in the first preparation image. If the scope of the design layout in the first preparation image is the same as the scope of the design layout in the first image, then the first image may be the same as the first preparation image.

[0033] The computer generates a first processed image by performing a first process on the first image (step S21b). In the first process in step S21b, the first parameter determined in step S11h is used.

[0034] The computer acquires a second image obtained by photographing the structure (step S21c). The second prepared image and the second image relate to the same structure. The extent of the structure shown in the second image may be the same as the extent of the structure shown in the second prepared image, or it may be different from the extent of the structure shown in the second prepared image.

[0035] The computer generates a second processed image by applying a second process to the second image (step S21d). In the second process in step S21d, the second parameter determined in step S11h is used.

[0036] The order of steps S21a to S21d can be changed as appropriate. For example, step S21c may be executed before step S21a, or steps S21b and S21d may be executed after steps S21a and S21c.

[0037] The computer sets the position of the other image in one of the first and second processed images in order to calculate the similarity (step S21e). For example, in the first processed image, a position is set where the similarity with the second processed image is calculated.

[0038] The computer performs template matching at the set location and calculates the similarity between the first processed image and the second processed image (step S21f).

[0039] The computer determines whether the entire range of either the first or second processed image has been explored by that point (step S21g). For example, if similarity has been calculated for all positions in the first processed image, it is determined that the entire range has been explored.

[0040] If it is determined that the entire range has not been explored, the computer changes the position of the other in one of the first and second processed images (step S21h). For example, in the first processed image, a position for which similarity has not yet been calculated is adopted.

[0041] After step S21h, step S21f is executed again. This recalculates the similarity at the modified position. Steps S21f to S21h are repeated until the entire range of one of the first and second processed images has been explored.

[0042] If it is determined that the entire range has been explored, the computer refers to the multiple similarity values ​​obtained and extracts the one with the highest similarity. The computer then refers to the position where that highest similarity was obtained. The computer determines that this position is the position where the first processed image and the second processed image fit together best (step S21i).

[0043] If an inspection image is also acquired when the second image is taken, that inspection image may be aligned with the first image. For example, if an abnormality in the structure is visible in the inspection image, the computer aligns the inspection image with the first image. As a result, the coordinates of the abnormality in the first image are calculated. The computer outputs the calculated coordinates (step S21j). This allows the computer user to easily understand at which coordinates in the design layout the abnormality is occurring.

[0044] The advantages of the embodiment will be explained. Traditionally, there is a technology that uses artificial intelligence (AI) to process images of design layouts in order to align them with photographs of structures. In this technology, the design layout image is processed by the AI ​​and an image similar to the photograph of the structure is output. Therefore, the photograph of the structure can be easily aligned with the image output by the AI. However, training the AI ​​requires a large amount of training data. Therefore, it has required a great deal of effort to prepare in advance.

[0045] In this embodiment of the present invention, instead of using AI, a first image of the design layout and a second image of the structure are processed separately. The first image is subjected to a first processing to generate a first processed image. The second image is subjected to a second processing to generate a second processed image. The first parameters used in the first processing and the second parameters used in the second processing are predetermined so that the first processed image and the second processed image are similar to each other. For example, if only one of the first or second image is processed, it is difficult to make the two images sufficiently similar. By processing the first and second images separately, it is possible to obtain a first processed image and a second processed image that are similar to each other. Then, alignment is performed using the first processed image and the second processed image. According to this embodiment, the image of the design layout and the photograph of the structure can be aligned more easily without using AI.

[0046] Preferably, at least one of the first and second processes includes a smoothing process. Generally, photographs contain image blur due to insufficient resolution. If the first process includes a smoothing process, blur is added to the image of the design layout, making the image of the design layout closer to a photograph. If the second process includes a smoothing process, random noise contained in the photograph is suppressed.

[0047] The size of the structures shown in the design layout may differ from the size of the structures shown in the photograph. Therefore, the second process may include resizing. Also, the orientation of the structures shown in the design layout may differ from the orientation of the structures shown in the photograph. Therefore, the second process may include rotation.

[0048] If both the first and second processes include a smoothing process, only the parameters of one of the smoothing processes may be optimized by step S10. For example, the first image is subjected to a smoothing process, and the second image is subjected to a smoothing process, resizing, and rotation. In this case, the parameters for the smoothing process of the first image, the parameters for the resizing process of the second image, and the parameters for the rotation of the second image may be optimized by step S10. The parameters for the smoothing process of the second image are set to arbitrary values ​​as appropriate.

[0049] (First variation) The design layout image may be generated by blending multiple images. Semiconductor devices generally have a multilayer structure consisting of multiple layers. Specifically, a semiconductor device includes a substrate, wiring layers, contact layers, etc. The substrate includes n-type or p-type semiconductor regions, insulating regions for electrically or spatially separating the semiconductor regions, etc. For example, multiple circuit elements such as diodes, capacitors, and transistors are formed on the substrate. Wiring layers are formed to electrically connect semiconductor regions to each other, and to electrically connect semiconductor regions to external circuits. Contact layers are formed to electrically connect wiring to semiconductor regions, or to electrically connect wiring to each other. Multiple wiring layers and multiple contact layers may be formed alternately in the thickness direction of the substrate.

[0050] When photographing structures, it is permissible to illuminate them with light. The wavelength of the light used can be selected as appropriate. In this case, depending on the combination of the light illumination position and wavelength, multiple layers of structure may be visible in the image.

[0051] As an example, infrared light is shone onto the semiconductor device from the substrate side. Infrared light is less easily absorbed by semiconductors than short-wavelength light such as green or blue light. Therefore, a photograph based on the reflected infrared light may include not only the structure of the substrate, but also the structure of the contact layer and the wiring layer. Furthermore, if the substrate is sufficiently thin, even with visible light, the photograph may include the structure of the substrate, the structure of the contact layer and the structure of the wiring layer. In such cases, an image showing the design layout of the substrate may be blended with an image showing the design layout of the contact layer or an image showing the design layout of the wiring layer.

[0052] As another example, visible light is shone on the semiconductor device from the wiring layer side. If the semiconductor device includes multiple wiring layers, and the insulating layers included in the wiring layers are transparent, light can be reflected by multiple wiring layers. A photograph based on the reflected visible light may include the structure of multiple wiring layers. In that case, the image of the design layout of the outermost wiring layer may be blended with images of the design layouts of the other wiring layers.

[0053] Figure 4 is a flowchart showing a part of the alignment method according to the first modified embodiment. In the first modified embodiment, step S10a shown in Figure 4 is performed instead of step S10 shown in Figure 2. First, the computer acquires a first partial image and a second partial image (step S12a). The first partial image shows the design layout of the first layer of the structure. The second partial image shows the design layout of the second layer of the structure. The range of the design layout in the first partial image and the range of the design layout in the second partial image are selected so that they overlap when viewed from the camera's shooting direction. For example, the range of the design layout in the first partial image and the range of the design layout in the second partial image overlap in the stacking direction of the structure.

[0054] The computer acquires a second preparation image (step S12b). The computer sets the first and second weights (step S12c). In the first modified example, the first preparation image is generated by blending the first and second partial images. The first weight is the weight parameter of the first partial image in the blend. The second weight is the weight parameter of the second partial image in the blend. In step S12c, the first and second weights are set to their initial values. The initial values ​​may be pre-set by the computer user or randomly set by the computer.

[0055] The computer generates a first preparation image by blending the first and second partial images (step S12d). At this time, the weights set immediately beforehand are used.

[0056] The computer calculates the similarity between the first and second preparation images (step S12e). Template matching can be used to calculate the similarity.

[0057] The computer determines whether the entire range of the first weight and the entire range of the second weight have been explored by that point (step S12f). For example, the range to be explored is predetermined for each of the first and second weights. The computer determines whether all of the predetermined ranges have been explored.

[0058] If it is determined that the entire range has not been explored, the computer changes either the first or second weight (step S12g). At this time, either only one of the first or second weight may be changed, or both the first and second weights may be changed.

[0059] After step S12g, step S12d is executed again. This generates the first preparation image again using the modified weights. The similarity between the newly generated first preparation image and the second preparation image is calculated. Steps S12d to S12g are repeated until the entire range of each weight has been explored.

[0060] Instead of searching the entire range, the number of loop executions may be set as the termination condition for the loop in steps S12d to S12g. A threshold for similarity may also be set as a termination condition. For example, if the similarity obtained in step S12e exceeds the threshold, the computer terminates the loop. The loop is repeated until the similarity exceeds the threshold or the number of loop executions reaches a predetermined value.

[0061] Furthermore, in steps S12f and S12g, optimization methods other than exhaustive search may be used. For example, in the initial stages of the search, the similarity is calculated while significantly changing at least one of the first and second weights. From the results, the computer calculates an approximation formula representing the relationship between the first weight and the similarity, and an approximation formula representing the relationship between the second weight and the similarity. From these approximation formulas, the computer determines the range in which the optimal first weight is estimated to be obtained, and the range in which the optimal second weight is estimated to be obtained. Within these ranges, the computer calculates the similarity while small changes are made to at least one of the first and second weights.

[0062] If it is determined in step S12f that the entire range has been explored, the computer refers to the multiple similarity values ​​obtained by repeating steps S12d to S12g. The computer extracts the highest similarity value from the multiple similarity values. The computer refers to the first and second weights that were used when the highest similarity value was obtained. The computer decides to use these first and second weights as the weights to be used in blending the images (step S12h).

[0063] After the weights are determined, the computer sets the first and second parameters (step S12i). The first and second parameters are values ​​used in the first and second processes, respectively. In step S12i, the first and second parameters are set to their initial values. These initial values ​​may be pre-set by the computer user or randomly set by the computer.

[0064] The computer performs a first processing on the first prepared image using a first parameter and a second processing on the second prepared image using a second parameter (step S12j). The computer calculates the similarity between the first prepared image that has undergone the first processing and the second prepared image that has undergone the second processing (step S12k). Template matching can be used to calculate the similarity.

[0065] The computer determines whether the entire range of the first parameter and the entire range of the second parameter have been explored by that point (step S12l). For example, the range to be explored is predetermined for each of the first and second parameters.

[0066] If it is determined that the entire range has not been explored, the computer changes either the first or second parameter (step S12m). At this time, either only one of the first or second parameter may be changed, or both the first and second parameters may be changed.

[0067] After step S12m, step S12j is executed again. This performs the first and second processes using the modified first and second parameters, and the similarity is recalculated. Steps S12j to S12m are repeated until the entire range of each parameter has been explored.

[0068] Instead of searching the entire range, the number of loop executions may be set as the termination condition for the loop in steps S12j to S12m. A threshold for similarity may also be set as a termination condition. For example, if the similarity obtained in step S12k exceeds the threshold, the computer terminates the loop. The loop is repeated until the similarity exceeds the threshold or the number of loop executions reaches a predetermined value.

[0069] If it is determined in step S12l that the entire range has been explored, the computer refers to the multiple similarity values ​​obtained by repeating steps S12j to S12m. The computer extracts the highest similarity value from the multiple similarity values. The computer refers to the first and second parameters that were used when the highest similarity value was obtained. The computer determines these first and second parameters as the parameters to be used in the first and second processing (step S12n).

[0070] In steps S12d to S12g, weights are set for each sub-image so that the blended first preparation image resembles a photograph of the structure. In steps S12j to S12m, the first and second parameters are determined so that the first preparation image, which has undergone the first processing, and the second preparation image, which has undergone the second processing, are similar.

[0071] Figure 5 is a flowchart showing a part of the alignment method according to the first modified embodiment. In the first modified example, step S20a shown in Figure 5 is performed instead of step S20 shown in Figure 3. In step S20a, the computer first acquires a third partial image (step S22a). The third partial image acquired in step S22a and the first partial image acquired in step S12a both show the design layout in the same first layer. The range of the design layout in the third partial image may be the same as or different from the range of the design layout in the first partial image.

[0072] The computer acquires the fourth partial image (step S22b). The fourth partial image acquired in step S22b and the second partial image acquired in step S12b both show the design layout in the same second layer. The range of the design layout in the fourth partial image may be the same as the range of the design layout in the second partial image, or it may be different from the range of the layout depicted in the second partial image. The range of the design layout in the third partial image and the range of the design layout in the fourth partial image are selected so that they overlap when viewed from the camera's shooting direction.

[0073] The computer generates the first image by blending the third and fourth sub-images (step S22c). At this time, the weight of the third sub-image is set to the first weight determined in step S12h. The weight of the fourth sub-image is set to the second weight determined in step S12h.

[0074] The computer generates a first processed image by performing a first process on the first image (step S22d). In the first process of step S22d, the first parameter determined in step S12n is used.

[0075] The computer acquires a second image obtained by photographing the structure (step S22e). The computer generates a second processed image by performing a second process on the second image (step S22f). In the second process in step S22f, the second parameter determined in step S12n is used.

[0076] The computer sets the position of the other image in one of the first and second processed images (step S22g). The computer performs template matching at the set position and calculates the similarity between the first and second processed images (step S22h).

[0077] The computer determines whether the entire range of one of the first and second processed images has been explored by that point (step S22i). If it is determined that the entire range has not been explored, the computer changes the position of the other of the first and second processed images (step S22j). After step S22j, step S22h is executed again. This recalculates the similarity at the changed position.

[0078] If it is determined that the entire range has been explored, the computer refers to the multiple similarity values ​​obtained and extracts the one with the highest similarity. The computer then refers to the position where that highest similarity was obtained. The computer determines that this position is the position where the first processed image and the second processed image fit together best (step S22k). After that, the computer may output the coordinates of the anomaly (step S22m).

[0079] According to the first modification, when a photograph includes a structure with multiple layers, the image of the design layout and the photograph of the structure can be aligned with greater accuracy.

[0080] (Second variation) In the first modified example, if the first or second process includes resizing, the amount of resizing change may be determined using a plurality of first and second prepared images. In the second modified example, the amount of resizing change is determined using a plurality of first and second prepared images.

[0081] Figure 6 is a flowchart showing a part of the alignment method according to a second modified embodiment. In the second modified embodiment, step S10b shown in Figure 6 is performed instead of step S10 shown in Figure 2. First, multiple pairs of first partial images and second partial images are prepared. The multiple pairs represent different areas of the same structure. In addition, a second preparation image corresponding to each pair is prepared.

[0082] The computer acquires multiple pairs (first partial image and second partial image) (step S13a). Next, the computer acquires multiple second preparation images corresponding to each of the multiple pairs (step S13b). This results in multiple combinations of the first partial image, the second partial image, and the second preparation images.

[0083] The computer selects a combination from several combinations to use for determining the weights (step S13c). The computer then performs the same process as in steps S12c to S12h shown in Figure 4 to determine the first and second weights (step S13d).

[0084] The computer uses the determined first and second weights to generate multiple first preparation images from the multiple pairs acquired in step S13a (step S13e). This results in multiple pairs of first and second preparation images.

[0085] The computer selects a pair of first and second preparation images to use in determining the resize change (step S13f). The computer sets the change amount (step S13g). In step S13g, the change amount is set to an initial value. The change amount may be pre-set by the computer user or may be set randomly by the computer.

[0086] The computer resizes either the first or second preparation image using a set amount of change (step S13h). Here, we describe an example where the second preparation image is resized. The computer calculates the similarity between the first preparation image and the resized second preparation image (step S13i). Template matching can be used to calculate the similarity.

[0087] The computer determines whether the entire range of the change amount has been explored (step S13j). If it determines that the entire range has not been explored, the computer changes the change amount (step S13k). After step S13k, step S13h is executed again. This causes the second preparation image to be resized again using the changed change amount. Steps S13h to S13k are repeated until the entire range of the change amount has been explored.

[0088] If it is determined in step S13j that the entire range has been explored, the computer refers to the multiple similarity values ​​obtained by repeating steps S13h to S13k. The computer selects the change amount that yields the highest similarity (step S13l).

[0089] The computer determines whether all pairs of first and second preparation images have been selected in the previous step S13f (step S13m). If there are any unselected pairs, step S13f is executed again. This selects any pairs that have not yet been selected. Using these pairs, steps S13g to S13k are executed again.

[0090] If all pairs are determined to have been selected in step S13m, the computer determines the change to be used for resizing based on the change amounts selected in the previous steps S13l (step S13n). For example, the computer determines the change amount that has been selected most frequently as the parameter to be used for resizing. The computer may also determine the average or median of the selected change amounts as the parameter to be used for resizing.

[0091] The parameter for the amount of change is susceptible to noise. When determining the amount of change using only a single pair of the first and second preparation images, noise may cause the determined amount of change to deviate from the optimal amount. According to the second modification, the amount of change is determined using multiple pairs. This allows for a more appropriate determination of the resize amount. By performing resizing using the determined amount of change, the first and second preparation images can be aligned with greater precision.

[0092] The horizontal and vertical changes in the image may be determined separately. In that case, in the first step S10b, the image is resized in either the vertical or horizontal direction. This determines the change in either the vertical or horizontal direction. Then, steps S13f to S13n are executed again. The image is resized in the other vertical or horizontal direction, and the change in the other vertical or horizontal direction is determined.

[0093] (Third variation) In the first modified example, if the first or second process involves rotation, the angle of rotation may be determined using a plurality of first and a plurality of second prepared images. In the third modified example, the angle of rotation is determined using a plurality of first and a plurality of second prepared images.

[0094] Figure 7 is a flowchart showing a part of the alignment method according to a third modified embodiment. In the third modified embodiment, step S10c shown in Figure 7 is performed instead of step S10 shown in Figure 2. The process in step S10c is generally the same as the process in step S10b.

[0095] First, the computer acquires multiple pairs (first and second partial images) (step S14a), similar to step S10b, and acquires multiple second preparation images corresponding to each pair (step S14b). The computer selects combinations to use for determining weights (step S14c) and determines the first and second weights (step S14d).

[0096] The computer generates multiple first preparation images from the multiple pairs acquired in step S14a (step S14e). The computer selects a pair of first and second preparation images to use for determining the rotation angle (step S14f). The computer sets the angle to an initial value (step S14g).

[0097] The computer rotates either the first or second preparation image by a set angle (step S14h). Here, we describe an example where the second preparation image is rotated. The computer calculates the similarity between the first preparation image and the rotated second preparation image (step S14i).

[0098] The computer determines whether the entire range of angles has been explored (step S14j). If it determines that the entire range has not been explored, the computer changes the angle (step S14k). After step S14k, step S14h is executed again. Steps S14h to S14k are repeated until the entire range of angles has been explored.

[0099] If the computer determines in step S14j that the entire range has been explored, it refers to the multiple similarities obtained by repeating steps S14h to S14k. The computer selects the angle with the highest similarity (step S14l).

[0100] The computer determines whether all pairs of first and second preparation images have been selected in the previous step S14f (step S14m). If there are any unselected pairs, step S14f is executed again. If it is determined that all pairs have been selected, the computer determines the angles to be used for rotation based on the angles selected in the previous steps S14l (step S14n). For example, the computer determines the angle that has been selected the most times as the parameter to be used for rotation. The computer may also determine the average or median of the selected angles as the parameter to be used for rotation.

[0101] The angle parameter is susceptible to noise. When determining the angle using only a single pair of the first and second preparation images, noise may cause the determined angle to deviate from the optimal angle. According to the third modification, multiple pairs are used to determine the angle used for rotation. This allows for a more appropriate determination of the rotation angle. By performing the rotation using the determined angle, the first and second preparation images can be aligned with greater precision.

[0102] If the first or second process includes both resizing and rotation, steps S10b and S10c are executed in combination. For example, after step S10b shown in Figure 6 is executed, steps S14f to S14n of step S10c are executed. This determines the amount of resizing change and the angle of rotation using multiple pairs. If the first and second processes include smoothing, after determining the amount of change and angle, the size of the smoothing filter may be searched using either pair of the first and second prepared images.

[0103] (Fourth variation) In each of the embodiments described above, images are aligned based on their similarity. The similarity is calculated using pixel values. That is, in the method described above, images are aligned at the pixel level. In the fourth modified embodiment, after pixel-level alignment, alignment is performed at a level smaller than a pixel (subpixel level).

[0104] Figure 8 is a flowchart showing a part of the alignment method according to the fourth modified embodiment. Step S25, shown in Figure 8, is performed after the alignment of the first and second processed images. When the process shown in Figure 3 is performed, step S25 is performed after step S21i. When the process shown in Figure 5 is performed, step S25 is performed after step S22k.

[0105] First, the computer enlarges the first processed image (step S25a) and then enlarges the second processed image (step S25b). The magnification of the second processed image is the same as that of the first processed image. The larger the magnification, the better the alignment accuracy at the subpixel level. For example, the magnification is preferably 2x or more, more preferably 5x or more, and most preferably 10x or more. Here, the enlarged first processed image is called the "first enlarged image." The enlarged second processed image is called the "second enlarged image."

[0106] The computer sets the positions of the first and second enlarged images in one of the first and second enlarged images (step S25c). The computer performs template matching at the set positions and calculates the similarity between the first and second enlarged images (step S25d).

[0107] The computer determines whether the entire range of one of the first and second enlarged images has been explored (step S25e). If it is determined that the entire range has not been explored, the computer changes the position of the other image in one of the first and second enlarged images (step S25f). After step S25f, step S25d is executed again. This recalculates the similarity at the changed position.

[0108] If the computer determines that the entire range has been explored, it refers to the multiple similarity values ​​obtained so far and extracts the one with the highest similarity. The computer then refers to the position where that highest similarity was obtained. The computer determines that this position is the position where the first and second enlarged images best match (step S25g). Subsequently, the coordinates after alignment are output.

[0109] According to step S25, the images are aligned at the subpixel level. This further improves the accuracy of the alignment. For example, if there is an abnormality in the inspection image, it is possible to obtain more precise coordinates of the abnormality in the design layout.

[0110] Furthermore, pixel-level alignment may be omitted, and sub-pixel-level alignment may be performed instead. For example, in the flowchart shown in Figure 3, steps S21e to S21i may be omitted, and steps S25a to S25g may be performed instead. When the positional misalignment between the first and second images is small, omitting pixel-level alignment can reduce the time required for alignment.

[0111] Figure 9 is a flowchart showing a part of the alignment method according to a second modified example of the embodiment. In subpixel-level alignment, the process shown in Figure 9 may be performed instead of the process shown in Figure 8. In step S26, the computer first generates a similarity map (step S26a). Multiple similarities obtained by the loop of steps S21f to S21h or the loop of steps S22h to S22j are used to generate the similarity map.

[0112] Figure 10 is a schematic diagram illustrating the process shown in Figure 9. Figure 10 is a schematic diagram showing a magnified portion of the first processed image. In the example shown in Figure 10, 25 pixels in the first processed image are shown. The numerical value for each pixel indicates the similarity when the position of the second processed image is set to that pixel. By repeatedly calculating the similarity, the pixel px 11 ~px 55 The similarity is calculated for each coordinate. In the illustrated example, the pixel px 33 And that's where the highest similarity score was obtained.

[0113] The computer uses the highest similarity and the similarity of the surrounding pixels to create the similarity map shown in Figure 10. The similarity map is divided into pixels px 11 ~px 55 This includes each coordinate and the similarity at each coordinate.

[0114] Next, the computer uses the multiple similarity values ​​included in the similarity map to calculate the location at the subpixel level where the similarity is maximized (step S26b). Gaussian fitting can be used to calculate the location. The distribution of similarity is approximately a normal distribution. Therefore, by using Gaussian fitting, the coordinates where the similarity is maximized (maximum) can be calculated. As a result, the pixel px 33 The computer calculates that the similarity is maximized at coordinate c1 within the given range. The computer outputs coordinate c1 as the position where the first processed image and the second processed image match best.

[0115] The computer may calculate the centroid position of multiple similarity factors instead of using Gaussian fitting. The computer outputs the centroid position as the position where the first processed image and the second processed image fit best.

[0116] (Examples) An embodiment of the alignment method according to the embodiment will be described. In the embodiment, the structure is a semiconductor device. The first process includes a smoothing process. The second process includes resizing and rotation.

[0117] Figure 11(a) is an example of the first partial image. Figure 11(b) is an example of the second partial image. Figure 11(c) is an example of the first preparation image. Figure 11(a) is image IMG1 showing the design layout of the substrate. Figure 11(b) is image IMG2 showing the design layout of the contact layer. The images in Figures 11(a) and 11(b) show the design layout in an area of ​​approximately 183 μm × 183 μm.

[0118] The computer performs steps S12c to S12h shown in Figure 4 to determine the weights for image IMG1 in Figure 11(a) and image IMG2 in Figure 11(b). The computer uses these weights to blend image IMG1 and image IMG2. As a result, the first prepared image IMG3 in Figure 11(c) is generated. In the illustrated example, the weight for image IMG1 is much larger than the weight for image IMG2. Therefore, image IMG3 in Figure 11(c) is similar to image IMG1 in Figure 11(a). For example, it can be seen that the shape of part A shown in Figure 11(a) has changed to the shape of part C shown in Figure 11(c) due to the blending.

[0119] The computer similarly blends multiple pairs of partial images other than those shown in Figures 11(a) and 11(b). This generates multiple first preparation images.

[0120] Figure 12 shows an example of a second preparation image. Image IMG4 in Figure 12 is acquired by photographing a semiconductor device in the manufacturing process from the substrate side using infrared light. The second preparation image IMG4 in Figure 12 is a photograph of an area of ​​approximately 183 μm × 183 μm.

[0121] The computer uses multiple pairs of first and second preparation images to determine the parameters for the first and second processes. The order in which the parameters for resizing, rotation, and smoothing are determined can be set as appropriate. In this embodiment, first, the amount of resizing change in the second process is determined. Next, the rotation angle in the second process is determined. Finally, the filter size for the smoothing process in the first process is determined.

[0122] First, the computer applies a smoothing process to the second prepared image. This smoothing process applies a smoothing filter to the second prepared image so that higher spatial frequencies specific to the optical image are excluded. In this embodiment, the size of the smoothing filter for the optical image is predetermined and is not explored.

[0123] Next, the computer resizes the second preparation image. Figure 13 is a graph showing the relationship between the amount of resizing change and the number of image pairs. When step S10b shown in Figure 6 is performed on the second preparation image, the result shown in Figure 13 is obtained. In Figure 13, the horizontal axis represents the increase or decrease in the number of pixels. The vertical axis represents the number of pairs that obtained the maximum similarity for each amount of change. The solid line shows the relationship between the amount of pixel change in the vertical direction and the number of pairs. The dashed line shows the relationship between the amount of pixel change in the horizontal direction and the number of pairs.

[0124] As shown in Figure 13, the largest number of pairs with the highest similarity were obtained at a resizing ratio of "0.999" in both the vertical and horizontal directions. Therefore, the computer determines that the vertical resizing amount of the second prepared image is "0.999". The computer also determines that the horizontal resizing amount of the second prepared image is "0.999".

[0125] Next, the computer rotates the resized second preparation image. Figure 14 is a graph showing the relationship between the rotation angle and the number of image pairs. When steps S14f to S14n of step S10c shown in Figure 7 are performed on the second preparation image, the result shown in Figure 14 is obtained. In Figure 14, the horizontal axis represents the rotation angle. The vertical axis represents the number of pairs that yielded the highest similarity at each angle.

[0126] As shown in Figure 14, the largest number of pairs with the highest similarity were obtained at a rotation angle of "-0.04 degrees". Therefore, the computer determines the rotation angle of the second preparation image to be "-0.04 degrees".

[0127] Finally, the computer applies a smoothing process to the first prepared image. Figure 15 is a graph showing the relationship between the size of the smoothing filter and the similarity. In Figure 15, the horizontal values ​​represent the size of the Gaussian filter in the X direction of the image. The vertical values ​​represent the size of the Gaussian filter in the Y direction of the image. Figure 15 shows the results of calculating the similarity for each combination of filter size in the X direction and filter size in the Y direction.

[0128] As shown in Figure 15, the similarity between the smoothed first prepared image and the resized and rotated second prepared image was maximized when the filter size in the X direction was "435 pixels" and the filter size in the Y direction was "389 pixels". Therefore, the computer determines the filter size in the X direction to be "435 pixels" and the filter size in the Y direction to be "389 pixels".

[0129] Through the above process, the filter size for the smoothing process in the first process is determined, and the amount of resizing change and the rotation angle in the second process are determined.

[0130] Next, the computer performs step S20a shown in Figure 5 using the determined parameters. Figure 16(a) is an example of a first processed image IMG5 after the first processing has been applied. Figure 16(b) is an example of a second processed image IMG6 after the second processing has been applied. The computer aligns the second processed image IMG6 shown in Figure 16(b) with the first processed image IMG5 shown in Figure 16(a).

[0131] Subsequently, the computer performs the process shown in Figure 8 or Figure 9. This aligns the second processed image IMG6 with the first processed image IMG5 at the subpixel level.

[0132] Figure 17 shows an example of an image illustrating the inspection results. The inspection image IMG7 shown in Figure 17 is obtained together with the original image from which the second processed image IMG6 was derived. Figure 17 shows the results of an electrical resistance inspection. In this inspection, the structure is scanned in the in-plane direction while applying voltage to the probe. Where the electrical resistance is low, current flows and is displayed in white. In the example shown in Figure 17, current is flowing in the upper left area. This area is determined to be abnormal.

[0133] The computer calculates the amount of movement required to align the second processed image with the first processed image. The computer resizes and rotates the inspection result image in the same way as the second image and aligns it with the first processed image. After that, the computer obtains the coordinates of the anomaly after alignment. If the coordinate system of the design layout image differs from the coordinate system of the inspection result image, the coordinate system of the inspection result image may be converted to the coordinate system of the design layout image. This outputs where the anomaly occurred in the design layout. Based on the inspection results, the anomaly can be easily identified.

[0134] The computer may output the position in the design layout image where the photograph of the structure best fits after alignment. The computer may also output the amount of resizing change, rotation angle, or smoothing filter size determined by the alignment method.

[0135] Figure 18 is a schematic diagram representing the configuration of a computer. The alignment method according to this embodiment is performed, for example, by the computer 90 shown in Figure 18. The computer 90 that performs the alignment method is used as an alignment device. The computer 90 includes a processing circuit (CPU or GPU) 91, ROM 92, RAM 93, storage device 94, input interface 95, output interface 96, and communication interface 97.

[0136] ROM92 stores programs that control the operation of computer 90. ROM92 contains the programs necessary for computer 90 to perform each of the processes described above. RAM93 functions as a memory area where the programs stored in ROM92 are loaded.

[0137] The CPU 91 includes processing circuits. The CPU 91 uses the RAM 93 as work memory and executes programs stored in at least one of the ROM 92 or the storage device 94. During program execution, the CPU 91 controls each component via the system bus 98 and performs various processes.

[0138] The memory device 94 stores data necessary for program execution and data obtained through program execution.

[0139] The input interface (I / F) 95 connects the computer 90 to the input device 95a. The input I / F 95 is, for example, a serial bus interface such as USB. The CPU 91 can read various data from the input device 95a via the input I / F 95.

[0140] The output interface (I / F) 96 can connect the computer 90 to the output device 96a. The output I / F 96 is a video output interface such as a Digital Visual Interface (DVI) or a High-Definition Multimedia Interface (HDMI®). The CPU 91 can transmit data to the output device 96a via the output I / F 96 and display an image on the output device 96a.

[0141] The communication interface (I / F) 97 allows the computer 90 to connect with a server 97a located outside the computer 90. The communication I / F 97 is, for example, a network card such as a LAN card. The CPU 91 can read various data from the server 97a via the communication I / F 97.

[0142] The storage device 94 includes one or more selected from Hard Disk Drives (HDDs) and Solid State Drives (SSDs). The input device 95a includes one or more selected from a mouse, keyboard, microphone (voice input), and touchpad. The output device 96a includes one or more selected from a monitor, projector, printer, and speaker. Devices that have the functions of both input device 95a and output device 96a, such as a touch panel, may also be used.

[0143] The alignment process may be performed by one computer 90 or by multiple computers 90.

[0144] The processing of the various data described above may be recorded as a program that can be executed by a computer on a magnetic disk (flexible disk and hard disk, etc.), an optical disk (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW, etc.), a semiconductor memory, or another non-transitory computer-readable storage medium.

[0145] For example, data on a recording medium is read by a computer (or embedded system). The recording format (storage format) on the recording medium is arbitrary. For example, a computer reads a program from the recording medium and causes the CPU to execute instructions based on this program. The acquisition (or reading) of the program by the computer may be done via a network.

[0146] According to the embodiments described above, a positioning method, positioning device, program, and storage medium are provided that enable more accurate positioning of an image of a design layout and a photograph of a structure.

[0147] In this specification, "or" indicates that "at least one" of the items listed in the text may be adopted.

[0148] Embodiments of the present invention include the following features. (Feature 1) On the computer, A first preparatory image showing the design layout of the structure and a second preparatory image obtained by photographing the structure are acquired. The first parameter used in the first processing and the second parameter used in the second processing are determined based on the similarity between the first prepared image that has undergone the first processing and the second prepared image that has undergone the second processing. By applying the first processing using the first parameter to the first image showing the design layout of the structure, a first processed image is obtained. The second image obtained by photographing the aforementioned structure is subjected to the second processing using the second parameter to obtain the second processed image. Based on the similarity between the first processed image and the second processed image, the first processed image and the second processed image are aligned. Alignment method. (Feature 2) At least one of the first and second processes includes a smoothing process, The alignment method according to Feature 1, wherein at least one of the first parameter and the second parameter includes the size of the smoothing filter. (Feature 3) The alignment method according to feature 1 or 2, wherein the first preparation image is generated by blending a first partial image showing the design layout of the first layer of the structure and a second partial image showing the design layout of the second layer of the structure. (Feature 4) The alignment method according to feature 3, wherein the computer determines a first weight for the first partial image and a second weight for the second partial image based on the similarity between the first prepared image and the second prepared image. (Feature 5) The first image is generated by blending a third partial image showing the design layout of the first layer of the structure and a fourth partial image showing the design layout of the second layer of the structure. The alignment method according to feature 4, wherein in the generation of the first image, the weight of the third partial image is set to the first weight, and the weight of the fourth partial image is set to the second weight. (Feature 6) At least one of the first and second processes includes resizing, The alignment method according to any one of features 1 to 5, wherein at least one of the first parameter and the second parameter includes the amount of change in resizing. (Feature 7) At least one of the first and second processes includes rotation, The alignment method according to any one of features 1 to 6, wherein at least one of the first parameter and the second parameter includes the angle of rotation. (Feature 8) At least one of the first and second processes includes resizing, Multiple pairs of the first partial image and the second partial image are obtained, For each pair, a plurality of first preparation images are generated by blending the first partial image and the second partial image using the first weight and the second weight. The alignment method according to feature 4, wherein the amount of resizing change is determined based on the similarity between the plurality of first prepared images and the plurality of second prepared images. (Feature 9) At least one of the first and second processes includes rotation, Multiple pairs of the first partial image and the second partial image are obtained, For each pair, a plurality of first preparation images are generated by blending the first partial image and the second partial image using the first weight and the second weight. The alignment method according to feature 4, wherein the angle of rotation is determined based on the similarity between the plurality of first prepared images and the plurality of second prepared images. (Feature 10) After the first processed image and the second processed image are aligned, the computer will perform the following: Based on the similarity between the enlarged first processed image and the enlarged second processed image, the enlarged first processed image and the enlarged second processed image are aligned at the subpixel level, or Using a similarity map based on the similarity between the first processed image and the second processed image, the first processed image and the second processed image are aligned at the subpixel level. The alignment method described in one of the features 1-9. (Feature 11) To the aforementioned computer, An inspection image showing the inspection results of the aforementioned structure is obtained, and the range captured in the inspection image is the same as the range captured in the second image. Based on the alignment result of the first processed image and the second processed image, the inspection image is aligned with the first processed image. The alignment method described in one of the features 1-10. (Feature 12) The alignment method according to feature 11, wherein if the inspection image shows an abnormal part of the structure, the inspection image is aligned with the first processed image, and then the computer outputs the coordinates of the abnormal part in the first image. (Feature 13) On the computer, By performing a smoothing process on the first image showing the design layout of the structure, the first processed image is obtained. By performing a smoothing process on the second image obtained by photographing the aforementioned structure, a second processed image is obtained. Based on the similarity between the first processed image and the second processed image, the first processed image and the second processed image are aligned. Alignment method. (Feature 14) The aforementioned structure is a semiconductor device, and the alignment method is as described in any one of features 1 to 13. (Feature 15) Equipped with a processing unit, An alignment device that performs the alignment method described in any one of the features 1 to 13. (Feature 16) A program that causes a computer to execute one of the alignment methods described in one of the features 1-13. (Feature 17) A storage medium containing the program described in Feature 16.

[0149] Although several embodiments of the present invention have been illustrated above, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. Furthermore, the embodiments described above can be implemented in combination with each other. [Explanation of symbols]

[0150] IMG1: First partial image, IMG2: Second partial image, IMG3: First preparation image, IMG4: Second preparation image, IMG5: First processed image, IMG6: Second processed image, IMG7: Inspection image

Claims

1. On the computer, A first preparation image showing the design layout of the structure and a second preparation image obtained by photographing the structure are acquired. The first parameter used in the first processing and the second parameter used in the second processing are determined based on the similarity between the first prepared image that has undergone the first processing and the second prepared image that has undergone the second processing. By applying the first processing using the first parameters to the first image showing the design layout of the structure, a first processed image is obtained. The second image obtained by photographing the aforementioned structure is subjected to the second processing using the second parameter to obtain the second processed image. Based on the similarity between the first processed image and the second processed image, the first processed image and the second processed image are aligned. Alignment method.

2. At least one of the first and second processes includes a smoothing process, The alignment method according to claim 1, wherein at least one of the first parameter and the second parameter includes the size of the smoothing filter.

3. The alignment method according to claim 1, wherein the first preparation image is generated by blending a first partial image showing the design layout of the first layer of the structure and a second partial image showing the design layout of the second layer of the structure.

4. The alignment method according to claim 3, wherein the computer determines a first weight for the first partial image and a second weight for the second partial image based on the similarity between the first prepared image and the second prepared image.

5. The first image is generated by blending a third partial image showing the design layout of the first layer of the structure and a fourth partial image showing the design layout of the second layer of the structure. The alignment method according to claim 4, wherein in the generation of the first image, the weight of the third partial image is set to the first weight, and the weight of the fourth partial image is set to the second weight.

6. At least one of the first and second processes includes resizing, The alignment method according to claim 1, wherein at least one of the first parameter and the second parameter includes the amount of change in resizing.

7. At least one of the first and second processes includes rotation, The alignment method according to claim 1, wherein at least one of the first parameter and the second parameter includes the angle of rotation.

8. At least one of the first and second processes includes resizing, Multiple pairs of the first partial image and the second partial image are acquired, For each pair, a plurality of first preparation images are generated by blending the first partial image and the second partial image using the first weight and the second weight. The alignment method according to claim 4, wherein the amount of change in resizing is determined based on the similarity between the plurality of first preparation images and the plurality of second preparation images.

9. At least one of the first and second processes includes rotation, Multiple pairs of the first partial image and the second partial image are acquired, For each pair, a plurality of first preparation images are generated by blending the first partial image and the second partial image using the first weight and the second weight. The alignment method according to claim 4, wherein the angle of rotation is determined based on the similarity between the plurality of first preparation images and the plurality of second preparation images.

10. After the first processed image and the second processed image are aligned, the computer will perform the following: Based on the similarity between the enlarged first processed image and the enlarged second processed image, the enlarged first processed image and the enlarged second processed image are aligned at the subpixel level, or The first processed image and the second processed image are aligned at the subpixel level using a similarity map based on the similarity between them. The alignment method according to claim 1.

11. To the aforementioned computer, An inspection image showing the inspection results of the aforementioned structure is obtained, and the range captured in the inspection image is the same as the range captured in the second image. Based on the alignment result of the first processed image and the second processed image, the inspection image is aligned with the first processed image. The alignment method according to claim 1.

12. The alignment method according to claim 11, wherein if an abnormal part of the structure is visible in the inspection image, the inspection image is aligned with the first processed image, and then the computer outputs the coordinates of the abnormal part in the first image.

13. The alignment method according to any one of claims 1 to 12, wherein the structure is a semiconductor device.

14. Equipped with a processing unit, An alignment device that performs the alignment method according to any one of claims 1 to 12.

15. A program that causes a computer to execute the alignment method described in any one of claims 1 to 12.

16. A storage medium storing the program described in claim 15.

Images