Method for evaluating quality of sem images
By combining Fourier transform and phase map reconstruction with large-scale and small-scale feature scoring, the problem of difficulty and inaccuracy in quality assessment caused by SEM image noise is solved, and accurate image quality assessment is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI PRECISION MEASUREMENT SEMICON TECH INC
- Filing Date
- 2022-12-15
- Publication Date
- 2026-06-16
AI Technical Summary
SEM images are noisy, and noise removal blurs other information, making it difficult to distinguish images of similar quality. Therefore, the image quality evaluation results in the existing technology are inaccurate.
The spectrogram and phase map are obtained by Fourier transform. The scores of large-scale features and small-scale features are combined to obtain the first and second scores respectively. The quality evaluation score of the SEM image is calculated by combining the scores.
It enables accurate evaluation of SEM image quality without noise removal, improves tolerance to image displacement and rotation, and solves the problems of difficulty and inaccuracy in image quality evaluation.
Smart Images
Figure CN115880259B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology, and in particular to a method for evaluating the quality of SEM images. Background Technology
[0002] In the field of integrated circuit inspection, to obtain higher-quality Scanning Electron Microscope (SEM) images, automatic evaluation of the acquired image quality is needed to assist in adjusting instrument parameters. Because SEM images are noisy, and noise removal blurs other information in the image, such as edge information, it becomes difficult to distinguish images of similar quality. Furthermore, when evaluating a set of images of the same scene, since SEM images are acquired in motion, there may be some displacement or rotation between the images. If only the edges of objects in the image are used to evaluate image quality, the image quality evaluation result will be inaccurate due to the differences in the edges contained within the images.
[0003] Therefore, it is necessary to propose a method for evaluating the quality of SEM images. Summary of the Invention
[0004] This invention provides a SEM image quality evaluation method to solve the technical problems in the prior art, such as the large noise in SEM images, which blurs other parts of the SEM image information while removing noise, making it difficult to distinguish images of similar quality and resulting in inaccurate image quality evaluation results.
[0005] In a first aspect, embodiments of the present invention provide a method for evaluating the quality of a SEM image, comprising: S1, acquiring a first SEM image and performing a Fourier transform on the first SEM image; S2, acquiring a spectrogram and a phase map corresponding to the first SEM image based on the result of the Fourier transform on the first SEM image; S3, acquiring a first score based on the amplitude of bright spots in the spectrogram, wherein the first score is a score for the large-scale features of the first SEM image; S4, reconstructing the SEM image based on the phase map to acquire a second SEM image; S5, acquiring a second score based on the edge information of the second SEM image, wherein the second score is a score for the small-scale features of the first SEM image; and S6, acquiring a quality evaluation score for the first SEM image based on the first score and the second score.
[0006] The beneficial effects are as follows: The SEM image quality evaluation method provided in this invention combines the richness of both large-scale and small-scale features to obtain a quality evaluation score for the SEM image. Large-scale features describe the main content of the SEM image, while small-scale features describe the object edge information. The resulting quality evaluation score, obtained by combining both, has a stronger tolerance for SEM image displacement and rotation. It enables accurate SEM image quality evaluation results without the need for image noise removal. In summary, this invention can solve the problem of difficult image quality evaluation due to high SEM image noise, and also addresses the problem of inaccurate image quality evaluation results caused by using only object edges in the image.
[0007] Optionally, step S3 includes: filtering out some bright spots that meet preset conditions in the spectrum, and obtaining the first score based on the amplitude of the bright spots that meet the preset conditions. The advantage is that the first score only needs to be obtained based on the amplitude of the bright spots that meet the preset conditions, thus minimizing computational load while ensuring the accuracy of the first score.
[0008] Optionally, step S3 includes: filtering high-brightness areas in the spectrum, where the high-brightness areas are those with the largest amplitude in the spectrum, and obtaining the first score based on the amplitude of the high-brightness areas. The beneficial effect is that high-brightness areas better reflect low-frequency signals in the spectrum and better reflect large-scale features in the image, resulting in more accurate quality evaluation scores.
[0009] Optionally, S3 includes: selecting high-brightness points with the largest amplitude and a preset number from the spectrum to form a point set L; and averaging the amplitudes of the selected high-brightness points to obtain the first score.
[0010] Optionally, S3 includes: when there are A first SEM images, selecting high-brightness points with the largest amplitude and a preset number from the spectrogram corresponding to one of the first SEM images to form a point set L1, where A is an integer greater than 1; the A first SEM images are SEM images used to evaluate the same scene but with different levels of clarity; using the lowest amplitude T1 in the point set L1 as a threshold, selecting high-brightness points with amplitudes greater than the threshold T1 from the spectrograms corresponding to the remaining images in the A first SEM images to form a point set Li corresponding to the i-th image in the A first SEM images, 1 < i ≤ A; averaging the amplitudes of the high-brightness points in the point set Li corresponding to the i-th image in the A first SEM images to obtain the first score corresponding to the i-th image in the first SEM images. Its beneficial effect is that when evaluating a group of SEM images of the same scene, since the SEM images are acquired in motion, there will be some displacement or rotation between the SEM images. If only the edges of objects in a SEM image are used to evaluate image quality, the image quality evaluation result will be inaccurate due to the differences in the edges contained in the SEM image. In this embodiment, the high-brightness screening method provided can improve the accuracy of image quality evaluation results for a set of SEM images of the same scene.
[0011] Optionally, in S3, the preset quantity is: or p is a preset value and makes Or make M×N represents the size of the first SEM image.
[0012] Alternatively, p∈(0,0.5).
[0013] Optionally, S5 includes: extracting edge information from the second SEM image to obtain a grayscale value distribution function of the edge points; calculating the negative of the average value of the grayscale values of the edge points to obtain a second score.
[0014] Optionally, via formula Calculate the negative of the average gray value of the edge points, where M×N is the size of the first SEM image and E is the gray value distribution function.
[0015] Optionally, step S4 includes: performing an inverse Fourier transform on the phase map to reconstruct the SEM image; step S6 includes: weighting the first score and the second score to obtain a weighted score, and using the weighted score as the quality evaluation score S of the first SEM image. Its beneficial effect is that by reconstructing the SEM image based on the phase map, the resulting second SEM image can highlight edge features and noise with drastic grayscale changes compared to the first SEM image, while removing areas with smooth grayscale changes and illumination variations, which helps improve the accuracy of small-scale feature acquisition. This embodiment of the invention can obtain an accurate quality evaluation score for the SEM image by setting corresponding weights under different needs; the higher the quality evaluation score S, the better the image quality. Attached Figure Description
[0016] Figure 1 A flowchart illustrating an embodiment of a SEM image quality evaluation method provided by this invention;
[0017] Figure 2 A schematic diagram of a spectrum diagram and phase diagram embodiment provided by the present invention;
[0018] Figure 3 This is a schematic diagram of a high-brightness distribution embodiment provided by the present invention;
[0019] Figure 4 This is a schematic diagram illustrating the changes in a phase diagram during processing, as provided in an embodiment of the present invention.
[0020] Figure 5 This is a schematic diagram illustrating an embodiment of the quality evaluation score distribution of different SEM images provided by the present invention;
[0021] Figure 6 This is a schematic diagram of an embodiment of a SEM image quality evaluation device provided by an embodiment of the present invention; Detailed Implementation
[0022] The technical solutions of the embodiments of this application are described below with reference to the accompanying drawings. In the description of the embodiments of this application, the terminology used in the following embodiments is for the purpose of describing specific embodiments only and is not intended to limit the application. As used in the specification and appended claims of this application, the singular expressions "a," "the," "the," "the," and "this" are intended to also include expressions such as "one or more," unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, "at least one" and "one or more" refer to one or more (including two). The term "and / or" is used to describe the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.
[0023] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized. The term "connection" includes direct connections and indirect connections, unless otherwise stated. "First" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated.
[0024] In the embodiments of this application, the words "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplarily" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of the words "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.
[0025] This invention provides a SEM image quality evaluation method and apparatus to solve the technical problem in the prior art that SEM images are noisy and that removing noise blurs information in other parts of the SEM image, making it difficult to distinguish images of similar quality (making image quality evaluation difficult), and to solve the problem that image quality evaluation results are inaccurate because only the edges of objects in the image are used to evaluate the image quality.
[0026] This invention provides a method for evaluating the quality of SEM images, the process of which is as follows: Figure 1 As shown, it includes:
[0027] S1. Obtain the first SEM image and perform a Fourier transform on the first SEM image;
[0028] S2. Based on the Fourier transform result of the first SEM image, obtain the spectrum and phase diagram corresponding to the first SEM image;
[0029] S3. Obtain a first score based on the amplitude of the bright spots in the spectrum, where the first score is the score of the large-scale features of the first SEM image.
[0030] S4. Reconstruct the SEM image based on the phase map to obtain a second SEM image;
[0031] S5. Based on the edge information of the second SEM image, obtain a second score, where the second score is the score of the small-scale features of the first SEM image;
[0032] S6. Obtain the quality evaluation score of the first SEM image based on the first score and the second score.
[0033] Specifically, the order in which the operations S3 and S4 need to be performed is not limited. That is, S3 and S4 can be performed simultaneously, or S3 can be performed first and then S4, or S4 can be performed first and then S3.
[0034] The beneficial effects are as follows: The SEM image quality evaluation method provided in this invention combines the richness of both large-scale and small-scale features to obtain a quality evaluation score for the SEM image. Large-scale features describe the main content of the SEM image, while small-scale features describe the object edge information. The resulting quality evaluation score, obtained by combining both, has a stronger tolerance for SEM image displacement and rotation. It enables accurate SEM image quality evaluation results without the need for image noise removal. This invention solves the problem of difficult image quality evaluation due to high SEM image noise, and also addresses the problem of inaccurate image quality evaluation results caused by using only object edges in the image.
[0035] Optionally, in S1, the steps include: the size of the first SEM image is M×N; obtaining the grayscale distribution function f(x,y) of the first SEM image, where (x,y) are the coordinates of a pixel, and f(x,y) is the grayscale value of the pixel at coordinates (x,y); and using the formula... Performing a Fourier transform on the first SEM image yields the frequency domain distribution function F(u,v); where u = 0, 1, 2, ..., M-1; v = 0, 1, 2, ..., N-1. Its beneficial effect lies in: through the formula... The Fourier transform result of the first SEM image was obtained, which is the frequency domain distribution function F(u,v) of the first SEM image.
[0036] Optionally, in S2, the complex expression of the frequency domain distribution function F(u,v) of the first SEM image is F(u,v)=a+bj, obtained through the formula... And P(u,v) = arctan(b,a), and obtain the spectrum Q(u,v) and phase map P(u,v) corresponding to the first SEM image respectively. The beneficial effect is that a spectrum map in which the low frequency part is located in the center of the image and the frequency gradually increases from the center to the edge is obtained.
[0037] In some embodiments, step S3 includes: filtering out some bright spots that meet preset conditions in the spectrum, and obtaining the first score based on the amplitude of the bright spots that meet the preset conditions. The advantage is that the first score only needs to be obtained based on the amplitude of the bright spots that meet the preset conditions, thus minimizing computational load while ensuring the accuracy of the first score.
[0038] In some embodiments, step S3 includes: filtering high-brightness areas in the spectrogram, where the high-brightness areas are those with the largest amplitude in the spectrogram, and obtaining the first score based on the amplitude of the high-brightness areas. The beneficial effect is that high-brightness areas better reflect low-frequency signals in the spectrogram and better reflect large-scale features in the image, resulting in more accurate quality evaluation scores.
[0039] In some embodiments, S3 includes: selecting high-brightness points with the largest amplitude and a preset number from the spectrum to form a point set L; and averaging the amplitudes of the selected high-brightness points to obtain the first score.
[0040] In some embodiments, S3 includes: when there are A first SEM images, selecting the one with the largest amplitude from the spectrum corresponding to any one of the first SEM images. A high-brightness points are formed into a point set L1 (equivalent to a point set Li, i = 1), where A is an integer greater than 1; A first SEM images are SEM images used to evaluate the same scene but with different levels of clarity; the lowest amplitude T1 in the point set L1 is used as a threshold, and high-brightness points with amplitudes greater than the threshold T1 are selected from the spectrograms corresponding to the remaining images in the A first SEM images to form a point set Li corresponding to the i-th image in the A first SEM images, 1 < i ≤ A, to obtain the large-scale features of the i-th image in the first SEM images; using the formula The average amplitude of the selected high-brightness images is calculated (where 1 ≤ i ≤ A) to obtain the first score S corresponding to the i-th image in the first SEM image. _Li The beneficial effect is that when evaluating a set of SEM images of the same scene, since SEM images are acquired in motion, there will be some displacement or rotation between them. If only the edges of objects in the SEM images are used to evaluate image quality, the image quality evaluation result will be inaccurate due to the differences in the edges contained in the SEM images. In this embodiment, the provided high-brightness screening method can improve the accuracy of image quality evaluation results for a set of SEM images of the same scene.
[0041] Optionally, in S3, the preset quantity is: or p is a preset value and makes Or make M×N represents the size of the first SEM image. (Using the formula...) or The one with the largest amplitude was selected or The first score S1 is obtained by averaging the amplitudes of the high-brightness points. Its beneficial effect is that, in this embodiment, the larger the first score S1, the richer the large-scale features. This indicates rounding down. This indicates rounding up to the nearest integer.
[0042] Alternatively, p∈(0,0.5).
[0043] In some embodiments, S4 includes: via formula The SEM image is reconstructed based on the phase map P(u,v) to obtain the grayscale distribution function f of the second SEM image. p(x,y). The SEM image is reconstructed based on the phase map P(u,v). The resulting second SEM image can highlight edge features and noise with drastic gray-level changes compared to the first SEM image, and remove areas with smooth gray-level changes and illumination changes, which is beneficial to improving the accuracy of small-scale feature acquisition.
[0044] In some embodiments, step S5 includes: extracting edge information from the second SEM image to obtain the grayscale distribution function E of the edge points; and using the formula... The second score S2 is obtained by calculating the negative of the average grayscale value of the edge points. Since SEM images are noisy, removing noise blurs information in other parts of the image, such as object edges, making it difficult to distinguish images of similar quality. This invention proposes a method to evaluate the richness of small-scale edge features in a phase-map reconstructed image without removing image noise. Compared to the first SEM image, the phase-map reconstructed image highlights small-scale edge features and noise with drastic grayscale changes, removing areas with smooth grayscale changes and illumination variations. When using edge detection algorithms to extract image edges, for blurry images, the extracted areas with drastic grayscale changes are high-frequency noise distributed throughout the image; for clear images, the grayscale changes of small-scale edge information are more drastic than noise, thus removing some noise. Therefore, the richness of small-scale edge features in an image can be distinguished by the number of extracted edges. The larger the second score S2, the richer the small-scale edge information in the image.
[0045] In some embodiments, step S6 includes: obtaining the quality evaluation score S of the first SEM image using the formula S = w1S1 + w2S2, where w1 and w2 are preset weights, and w1 + w2 = 1. The beneficial effect is that the embodiments of the present invention can obtain an accurate quality evaluation score for the SEM image by setting corresponding weights under different needs; the higher the quality evaluation score S, the better the image quality.
[0046] To illustrate the invention in this application in more detail, specific examples are provided below:
[0047] A SEM image to be evaluated is obtained, referring to the first SEM image mentioned above. The size of the SEM image to be evaluated is M×N, where M and N are the length and width of the SEM image to be evaluated, respectively. A Fourier transform is performed on the gray-level distribution function f(x,y) of the SEM image to be evaluated to obtain the frequency distribution function F(u,v) of the SEM image to be evaluated. Where u = 0, 1, 2, ..., M-1; v = 0, 1, 2, ..., N-1; the frequency distribution function F(u,v) is calculated to obtain the spectrum Q(u,v) and phase diagram P(u,v). The spectrum Q(u,v) is as follows: Figure 2 As shown in (1), the phase diagram P(u,v) is as follows Figure 2 As shown in (2) in the figure, where, as Figure 2 The spectrum diagram shown is the transformed image. The transformation involves shifting the low-frequency components to the image center, with the frequency gradually increasing from the center to the image edges. Assume the complex expression of the frequency distribution function F(u,v) is a+bj, and Q(u,v) and P(u,v) are calculated as follows: P(u,v)=arctan(b,a).
[0048] Large-scale features of the SEM image to be evaluated are extracted from the spectrogram Q(u,v) to obtain a first score S1 for evaluating the richness of large-scale features. Bright spots in the spectrogram indicate high energy at the corresponding frequency, meaning the features they correspond to have high repeatability in the time-domain image. These bright spots, such as the high bright spots mentioned earlier, are mainly concentrated in low-frequency signals, corresponding to large-scale features in the image. Figure 2 The high-brightness points in the spectrogram shown in (1) are concentrated at the center of the image. The spectrogram is then filtered to extract the points with the largest amplitudes. By identifying high-brightness points, p∈(0,0.5), large-scale features in the time-domain image can be obtained. In this embodiment, p=0.001, the largest amplitude is selected from the spectrogram. A set of points L consists of several high-brightness points, and the selected points are as follows: Figure 3 As shown. The average of the amplitudes of the selected points is used to obtain the score S1 for the richness of large-scale features of the image: The larger the S1 value, the richer the large-scale features in the image.
[0049] When evaluating the quality of a set of SEM images with varying degrees of sharpness in the same scene, another method for selecting high-brightness sets can be used due to prior information about the same shooting scene. First, select any image from this set of SEM images and filter its high-brightness set using the method described above, taking the lowest amplitude T1 in the high-brightness set as a threshold. Then, use the threshold T1 to filter out high-brightness points in each image with amplitudes greater than the threshold T1, forming a high-brightness set. Calculate the Sb value for each image in this set of SEM images. 1i .
[0050] The image is reconstructed using the phase map P(u,v), i.e., by performing an inverse Fourier transform on P(u,v), as shown in the following formula: exist Figure 4In (1), original image 401 and original image 406 correspond to different SEM images, that is, two different first SEM images. The images reconstructed from their phase images (not shown) are as follows: Figure 4 As shown in SEM images 402 and 405, it can be seen that the reconstructed image obtained by the method provided in this embodiment of the invention highlights small-scale edge features and noise with drastic gray-level changes compared to the phase image, while removing areas with smooth gray-level changes and illumination variations. Then, an edge detection algorithm is used to extract edges from SEM images 402 and 405 respectively to obtain edge distribution images 403 and 404. The gray-level distribution function, E, of the edge points in edge distribution images 403 and 404 is obtained respectively. By calculating the negative of the average gray-level value of the edge points, a second score S2 for small-scale features is obtained, calculated using the following formula:
[0051] Because noise is uniformly distributed throughout the SEM image, such as Figure 4 As shown in edge distribution images 403 and 404, when the image is blurred, the extracted parts with drastic gray-level changes are noise; when the image is clear, the gray-level changes of small-scale edge information are more drastic than noise, thus removing some noise. Therefore, the formula for calculating the second score S2 is the negative of the average gray-level values of the edge points. The larger S2 is, the more small-scale edge information is in the image. There are many edge detection algorithms to choose from, including Sobel, Laplacian, Prewitt, and Canny. Among them, the Canny algorithm is a preferred algorithm because it removes more noise when using double thresholding to suppress false edges.
[0052] The quality evaluation score S considers both large-scale features S1 and small-scale features S2, providing a more comprehensive assessment of the image's feature richness. The quality evaluation score S of the SEM image is calculated as S = w1S1 + w2S2, where w1 and w2 are weights, and w1 + w2 = 1. w1 and w2 are determined by measuring the proportion of large-scale and small-scale features contained in the image within the usage scenario; a higher proportion carries a higher weight. In this embodiment, w1 = w2 = 0.5. A higher S indicates better image quality. When six SEM images exist, these six SEM images are respectively... Figure 5SEM images 501, 502, 503, 504, 505, and 506 are shown in the table. Using the above method, the first score S1 of SEM image 501 is 19.17, the second score S2 is -92.83, and the quality evaluation score S is -73.66; the first score S1 of SEM image 502 is 20.62, the second score S2 is -91.73, and the quality evaluation score S is -71.11; the first score S1 of SEM image 503 is... The first score for SEM image 504 was 20.13, the second score S2 was -91.56, and the quality evaluation score S was -71.43; the first score for SEM image 505 was 22.1, the second score S2 was -77.6, and the quality evaluation score S was -55.5; the first score for SEM image 506 was 26.61, the second score S2 was -53.55, and the quality evaluation score S was -26.93. SEM image 506 is the clearest image, and its quality evaluation score S is also the highest. The obtained quality evaluation score S is consistent with the actual situation.
[0053] Based on the SEM image quality evaluation method described in any of the above embodiments, this invention provides a SEM image quality evaluation device for performing the SEM image quality evaluation method as described in any of the above embodiments. The SEM image quality evaluation device is as follows: Figure 6 As shown, it includes: a first acquisition unit 601, a second acquisition unit 602, a third acquisition unit 603, a fourth acquisition unit 604, a fifth acquisition unit 605, and a sixth acquisition unit 606; the first acquisition unit 601 is used to acquire a first SEM image and perform a Fourier transform on the first SEM image; the second acquisition unit 602 is used to acquire a spectrogram and a phase map corresponding to the first SEM image based on the result of the Fourier transform on the first SEM image; the third acquisition unit 603 is used to acquire a first score based on the amplitude of the bright spots in the spectrogram, the first score being a score of the large-scale features of the first SEM image; the fourth acquisition unit 604 is used to reconstruct the SEM image based on the phase map to acquire a second SEM image; the fifth acquisition unit 605 is used to acquire a second score based on the edge information of the second SEM image, the second score being a score of the small-scale features of the first SEM image; the sixth acquisition unit 606 is used to acquire a quality evaluation score of the first SEM image based on the first score and the second score.
[0054] The beneficial effects are as follows: The SEM image quality evaluation device provided in this invention obtains a SEM image quality evaluation score by simultaneously combining the richness of large-scale and small-scale features. Large-scale features are used to describe the main content of the SEM image, while small-scale features are used to describe the object edge information of the SEM image. The SEM image quality evaluation score obtained by combining the two has a stronger tolerance for the displacement and rotation of the SEM image. It can achieve accurate SEM image quality evaluation results without removing image noise. In summary, this invention can solve the problem of difficult image quality evaluation caused by high noise in SEM images, and the problem of inaccurate image quality evaluation results caused by using only the edges of objects in the image to evaluate image quality.
[0055] All relevant content of each step involved in the above method embodiments can be referenced from the functional description of the corresponding unit module, and will not be repeated here.
[0056] The above description is merely a specific implementation of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of this application should be covered within the protection scope of the embodiments of this application. Therefore, the protection scope of the embodiments of this application should be determined by the protection scope of the claims.
Claims
1. A method for evaluating the quality of SEM images, characterized in that, include: S1. Obtain the first SEM image and perform a Fourier transform on the first SEM image; S2. Based on the Fourier transform result of the first SEM image, obtain the spectrum and phase diagram corresponding to the first SEM image; S3. Obtain a first score based on the amplitude of the bright spots in the spectrum, where the first score is the score of the large-scale features of the first SEM image. S4. Reconstruct the SEM image based on the phase map to obtain a second SEM image; S5. Based on the edge information of the second SEM image, obtain a second score, where the second score is the score of the small-scale features of the first SEM image; S6. Obtain the quality evaluation score of the first SEM image based on the first score and the second score; the SEM image is a scanning electron microscope image.
2. The SEM image quality assessment method according to claim 1, characterized in that, In S3, the following are included: Select some bright spots that meet preset conditions in the spectrum, and obtain the first score based on the amplitude of the bright spots that meet the preset conditions.
3. The SEM image quality evaluation method according to claim 2, characterized in that, In S3, the following are included: High-brightness points are selected from the spectrum graph, and the high-brightness points are the bright spots with the largest amplitude in the spectrum graph. The first score is obtained based on the amplitude of the high-brightness points.
4. The SEM image quality evaluation method according to claim 3, characterized in that, In S3, the following are included: High-brightness points with the largest amplitude and a preset number are selected from the spectrum to form a point set. ; The average amplitude of the selected high-brightness points is used to obtain the first score.
5. The SEM image quality evaluation method according to claim 3, characterized in that, In S3, the following are included: When there are A first SEM images, high-brightness points with the largest amplitude and a preset number are selected from the spectrum corresponding to one of the first SEM images to form a point set. A is an integer greater than 1; A first SEM images are SEM images used to evaluate the same scene but with different levels of clarity; The point set The lowest amplitude As a threshold, in the spectral maps corresponding to the remaining images in the first SEM image A, images with amplitudes greater than the threshold are selected respectively. The high-brightness points form the point set corresponding to the i-th image in A of the first SEM images. , 1 < i ≤ A; The point set corresponding to the i-th image in A of the first SEM images The average amplitude of the high-brightness points in the first SEM image is calculated to obtain the first score corresponding to the i-th image in the first SEM image.
6. The SEM image quality evaluation method according to claim 4 or 5, characterized in that, In S3, the preset quantity is: or p is a preset value and makes ≥2, or make ≥2, The size of the first SEM image.
7. The SEM image quality evaluation method according to claim 6, characterized in that, p∈(0,0.5)。 8. The SEM image quality assessment method according to claim 1, characterized in that, S5 includes: extracting edge information from the second SEM image to obtain the gray value distribution function of the edge points; Calculate the negative of the average gray value of the edge points to obtain the second score.
9. The SEM image quality evaluation method according to claim 8, characterized in that, Through formula Calculate the negative of the average gray value of the edge points. Let E be the size of the first SEM image, and E be the grayscale distribution function.
10. The SEM image quality assessment method according to claim 1, characterized in that, S4 includes: performing an inverse Fourier transform on the phase map to reconstruct the SEM image; In step S6, the following steps are included: weighting the first score and the second score to obtain a weighted score, and using the weighted score as the quality evaluation score of the first SEM image. .