A method for repairing defocus in wafer imaging
By moving the standard sample in the XY plane, a mapping relationship between the defocus radius and the distance is established. The Wiener filtering algorithm is used to repair the defocus in wafer imaging, which solves the blurring problem caused by the position deviating from the depth of field during wafer imaging, and improves the detection accuracy and imaging quality.
Patent Information
- Application Number
- CN202410910419.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-09
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2044-07-09
AI Technical Summary
Defocusing and blurring caused by the wafer's position deviating from the depth of field during wafer imaging affects detection accuracy and quality.
By moving the standard sample in the XY plane, taking multiple images and recording the distance between the camera and the standard sample in the Z direction, a mapping relationship between the defocus radius and the distance is established. The Wiener filtering algorithm is then used to repair the defocused image and restore the clarity of the wafer image.
This technology improves image quality and detection accuracy, reduces resource waste, and increases image utilization without requiring additional camera displacement components.
Abstract
Description
Technical Field
[0001] This invention belongs to the field of wafer inspection technology, and relates to wafer imaging repair, specifically to a method for wafer imaging defocus repair. Background Technology
[0002] Wafer image inspection is a crucial step in semiconductor manufacturing, used to identify various defects that may occur during wafer fabrication, such as microparticles, scratches, and lattice misalignments. Image inspection technology allows these minute defects to be detected without contacting the wafer, thereby improving the overall quality and performance of the product.
[0003] Because the features to be inspected on a wafer are extremely small, wafer image inspection typically requires an imaging system with high magnification to be used close to the wafer to observe the details of the microstructure. Due to the close shooting distance, combined with the small depth of field of the high-magnification imaging system, the position of the wafer's surface to be photographed needs to be within a small area.
[0004] In the corresponding wafer image inspection equipment, due to factors such as the tilt of the wafer stage, the tilt of the wafer, the uneven thickness of the wafer itself, and the slight deformation of the wafer caused by fixing the wafer, the photographed surface of the wafer is often located at least part of the depth of field of the imaging system, while the position of the camera itself is usually fixed. This will cause the image to be out of focus and blurry, interfering with the wafer inspection.
[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of the present invention, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0006] To address the shortcomings of existing technologies that easily lead to defocusing and blurring in imaging images, this invention provides a method for repairing defocusing in wafer imaging. The method comprises: moving a standard sample within the XY plane and capturing n images of the standard sample at different positions, obtaining images respectively. i And record the corresponding distance Dis between the camera and the standard in the Z direction. i Where i = 1, 2, 3...n; find the clearest image. a and the corresponding Dis a For each image i The sharpness of the restored out-of-focus image is calculated to be closest to that of the original image. a defocus radius R i Calculate ΔDis i =Dis a -Dis i Fitting R iWith ΔDis i Obtain the first mapping relationship; move the wafer to the preset position and capture an image of the wafer. b And record the corresponding distance Dis b Calculate ΔDis b =Dis a -Dis b According to ΔDis b The first mapping relationship yields the corresponding defocus radius R. b Using R b For Image b The image is restored to its original state, resulting in a wafer imaging image after the defocused image has been repaired.
[0007] According to another aspect of the embodiments of this application, feature points are provided on the standard sample, and each image... i It contains at least 3 feature points.
[0008] According to another aspect of the embodiments of this application, each image i In the given information, there are at least three non-collinear feature points.
[0009] According to another aspect of the embodiments of this application, the defocus radius R is obtained in the following manner. i Estimated defocus radius R irec Using R irec Perform out-of-focus image restoration to obtain the restored image. irec Comparison of images irec With Image a When the difference in sharpness between the two meets the preset conditions, the corresponding defocus radius R is used. irec As the defocus radius R i .
[0010] According to another aspect of the embodiments of this application, the images are compared in the following manner. irec With Image a Clarity: for image irec Image a Edge detection and the mean gradient of the image are performed sequentially to obtain the image. irec Image a The corresponding gradient mean G irec G a Calculate G irec With G a The squared error value is considered to meet the preset condition when the squared error value is less than the preset threshold.
[0011] According to another aspect of the embodiments of this application, the estimated defocus radius R is obtained by traversing a preset range with a fixed step size. irec .
[0012] According to another aspect of the embodiments of this application, the Wiener filtering algorithm is used for defocus image restoration.
[0013] According to another aspect of the embodiments of this application, the method further includes: moving the wafer to multiple preset positions and capturing multiple wafer images. b The wafer imaging images after defocus image restoration were calculated and obtained respectively.
[0014] According to another aspect of the embodiments of this application, in capturing an image i At this time, the object being photographed forms an angle with the XY plane.
[0015] According to another aspect of the embodiments of this application, the angle between the photographed surface of the standard sample and the XY plane is not less than the angle between the photographed surface of the wafer and the XY plane.
[0016] The present invention has at least the following beneficial effects: It can obtain the relationship between the defocus radius parameters involved in image restoration and the distance between the camera and the subject through a single detection. When imaging the wafer subsequently, only the distance relationship between the camera and the subject needs to be considered to obtain a better defocus restoration effect, thereby achieving defocus restoration of the wafer image; it eliminates the need for additional camera displacement components, making implementation convenient and cost-effective; and it eliminates the need for repeated analysis of each wafer, not only improving image utilization and reducing resource waste caused by reshooting, but also significantly improving imaging quality and detection accuracy. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are some embodiments of this application, but not all embodiments.
[0018] This invention provides a defocus restoration method for defocused wafer images, aiming to restore the image sharpness of the wafer through precise control and calculation, and to optimize the image quality by effectively determining and applying the defocus radius.
[0019] First, the relationship between the defocus radius and the distance between the camera and the subject is obtained to find optimal defocus radius parameters and perform image restoration. Specifically, the standard sample is moved in the XY plane, and n images of the standard sample are taken at different positions to obtain images respectively. i And record the corresponding distance Dis between the camera and the standard in the Z direction. iWhere i = 1, 2, 3...n. During this process, the camera's position remains fixed, especially its Z-axis position.
[0020] A standard sample can be a specific object used to obtain the relationship between the defocus radius and the distance between the camera and the subject, or it can be a wafer. More specifically, it can be an object with a specific pattern drawn on its surface, which the camera can then capture for further image analysis; this specific pattern can be a point, a straight line at an angle, or a specific pattern. Specifically, when feature points are set on the standard sample, the density of these feature points can be set so that each image... i It contains at least 3 feature points. In some cases, in order to further clarify the image... i The actual sharpness can make each image clearer. i In the given information, there are at least three non-collinear feature points.
[0021] When using a wafer as a standard, the wafer needs to have observable features, such as grain separators and the structure of the grains themselves. When using a non-wafer object as a standard, to broaden the applicability of the method of the present invention, the size of the standard can be matched to the size of the wafer. For example, the size of the standard can be the same as some commonly used wafer sizes, such as 2-inch, 3-inch, 4-inch, ..., 18-inch wafers; the size of the standard can also be larger than the size of the wafer. For example, when performing image inspection on an 18-inch wafer, the size of the standard can be slightly larger than 18 inches; other shapes of standard can also be used, such as square, circular, or other shapes.
[0022] Regarding the distance Dis between the camera and the standard in the Z direction i The distance can be measured using a ranging device. Specifically, a ranging device can be set up near the camera to detect the distance between the standard sample or wafer and the ranging device. Then, based on the positional relationship between the ranging device and the camera, the distance between the standard sample or wafer and the camera can be calculated.
[0023] After obtaining multiple images i and the corresponding Dis i Then, find the clearest image. a and the corresponding Dis aImage sharpness can be assessed through edge detection, frequency domain analysis, statistical methods, or even subjective evaluation. Edge detection is a relatively quantifiable method, and more specifically, sharpness can be quantified using gradient analysis. A more convenient method is subjective evaluation by humans to select images with better sharpness.
[0024] Select the clearest image. a Then, for a single image i Multiple defocus radii were used as input parameters for the defocus restoration model to perform defocus restoration. The image with the sharpness closest to the aforementioned image was then obtained. a Then, the corresponding defocus radius parameter R is obtained. i This is used as the optimal parameter for out-of-focus restoration of the image. This step is performed on multiple images separately to obtain the desired image for each image. i The corresponding defocus radius R i Combining various images i The corresponding Dis i Calculate the distance Dis between each image and the sharpest image. i The difference between them, that is, it can be calculated by ΔDis i =Dis a -Dis i To quantify this difference. Then, R is fitted. i With ΔDis i To obtain the first mapping relationship.
[0025] The first mapping relationship mainly refers to the mapping between the change in distance between the subject and the camera and the parameters required for the optimal defocus restoration model. For the defocus restoration model, the defocus radius is a core parameter, and it can be used to construct the corresponding model. Specifically, Wiener filtering is typically used in defocus restoration models. Correspondingly, a point spread function can be constructed using the defocus radius to control the defocus restoration effect of the Wiener filtering algorithm, thus enabling the use of the Wiener filtering algorithm for defocus restoration.
[0026] Subsequently, the wafer was photographed, and the aforementioned model was used for image defocusing correction. Specifically, the wafer was moved to a preset position, and an image of the wafer was captured. b And record the corresponding distance Dis bThe preset position can be either the same position as when the standard sample was photographed, or a preset position set according to the testing requirements. There is usually at least one preset position. When multiple preset positions are set, the setting of the preset position does not need to refer to the position when the standard sample image was photographed. After obtaining the image... b and the corresponding Dis b Then, calculate ΔDis. b =Dis a -Dis b According to ΔDis b The first mapping relationship yields the corresponding defocus radius R. b Using R b For Image b The image is restored to its original state, resulting in a wafer imaging image after the defocused image has been repaired.
[0027] Similarly, in the case of Image b When performing defocus restoration, Wiener filtering can usually be selected as the defocus restoration model. Accordingly, a point spread function can be constructed using the defocus radius to control the defocus restoration effect of the Wiener filtering algorithm, so as to use the Wiener filtering algorithm for defocus restoration.
[0028] In the aforementioned process, a single detection can obtain the relationship between the defocus radius parameters involved in image restoration and the distance between the camera and the subject. When imaging the wafer subsequently, only the distance relationship between the camera and the subject needs to be considered to obtain a good defocus restoration effect, thereby achieving defocus restoration of wafer imaging. No additional camera displacement components are required; existing wafer imaging equipment can be used directly to achieve good imaging results, making implementation convenient and cost-effective. Each wafer does not need to be analyzed again, which not only improves image utilization and reduces resource waste caused by reshooting, but also significantly improves imaging quality and detection accuracy.
[0029] In response to a single image i Calculate the optimal defocus radius parameter R i Specifically, the defocus radius R can be obtained in the following way: i For a single image i First, estimate a defocus radius R. irec1 Using R irec1 For image i Perform out-of-focus image restoration to obtain the restored image. irec1 Comparison images irec1 With Image a To adjust the focus radius to R, when the difference in focus between the two is significant, i.e., the image restoration effect is not good enough. irec2Using R irec2 For image i Perform out-of-focus image restoration to obtain the restored image. irec2 ; re-image the repaired image irec2 With Image a Compare sharpness; repeat multiple times until the repaired image matches the original image. a When the difference in sharpness meets the preset conditions, the corresponding defocus radius R is used. irec As the defocus radius R i .
[0030] When estimating the defocus radius R irec In some cases, estimations can be made based on experience; alternatively, certain rules can be set for the defocus radius R. irec The feasible range of values is traversed. Regarding the traversal method, when traversing discrete points within a continuous range, the step size can be set based on the computation time or efficiency. When traversing a pre-defined set of possible values, the traversal can proceed sequentially. During the traversal, if the repaired sharpness difference meets preset conditions, the traversal can either end directly or continue to find a better defocus radius.
[0031] Comparing images irec With Image a When assessing image sharpness, it can be determined either through manual evaluation or through quantitative analysis. When using quantitative analysis, the same image sharpness analysis algorithm can be used to analyze different parts of the image. irec Image a The sharpness values are compared to obtain the difference in sharpness between the two images. As a feasible implementation method, the images can be processed sequentially. irec Image a Perform edge detection and calculate the mean gradient of the image to obtain the image. irec Image a The corresponding gradient mean G irec G a Compare the gradient mean G irec G a The difference between the two is considered to be similar in sharpness if the difference falls within a preset range, thus meeting the corresponding conditions. Specifically, G can be calculated. irec With G aThe squared error value is used to determine whether the sharpness difference meets the preset condition. When the squared error value is less than the preset threshold, the sharpness difference is considered to meet the preset condition. The preset threshold can be set as needed. For example, when it is necessary to minimize the sharpness difference of the restored image, a smaller preset threshold can be set; when a faster processing speed is required, a relatively larger preset threshold can be set.
[0032] Furthermore, images can be captured from multiple locations on the wafer, and then combined from these images to form a larger image of the wafer. Specifically, the wafer can be moved to multiple preset positions and multiple wafer images can be captured. b The wafer imaging images after defocus image restoration are calculated and obtained separately, and these images can then be further combined and calculated.
[0033] As can be seen from the foregoing, one of the key aspects of this invention lies in obtaining the first mapping relationship. Accordingly, this can be achieved by increasing Dis when constructing the mapping relationship. i The range is determined so that the resulting first mapping remains valid over a larger scope. Accordingly, this can be achieved by capturing an image. i At this time, the object being photographed is made to form an angle with the XY plane to increase the Dis. i The range of variation.
[0034] Accordingly, to ensure the validity of the first mapping relationship, the angle between the photographed surface of the standard sample and the XY plane is not less than the angle between the photographed surface of the wafer and the XY plane, so that the Dis corresponding to the photographed surface of the wafer... b The range is within the Dis area corresponding to the photographed surface of the standard sample. i Within the scope.
[0035] The following is a further explanation of the content of the present invention through a feasible implementation process.
[0036] S0: A calibration plate, similar in size and shape to the wafer to be photographed, is placed on the displacement stage as a standard sample, and the standard sample is slightly tilted to form an angle between its photographed surface and the XY plane; the calibration plate is provided with feature points, and the feature points have a certain density so that there are at least 3 non-collinear feature points in each photographed area of the camera; the displacement stage can drive the calibration plate to move approximately along the XY plane; the distance sensor and the camera are fixedly set, the distance sensor detects the distance between the imaging area of the camera and the distance sensor, and then calculates the distance between the camera and the imaging area of the camera based on the positional relationship between the distance sensor and the camera, so as to ensure that the distance between the camera and the standard sample in the Z direction can be obtained.
[0037] S1: Referring to the shape of the standard sample, at least 9 preset shooting positions are arrayed in the projection area of the sample's photographed surface; the standard sample is moved by a displacement stage so that the imaging area of the camera is located in the aforementioned preset positions, and images Image1 to Image9 are captured respectively, and the corresponding distances between the camera and the standard sample in the Z direction Dis1 to Dis9 are recorded.
[0038] S2: Use the Sobel operator on the image i Perform edge detection, and then calculate the image. i gradient mean G i Choose the gradient mean G i The largest image is the one with the best clarity, thus yielding the clearest image. a and the corresponding Dis a G a .
[0039] S3: For images Image1 to Image9, set a preset threshold A and perform the following operations respectively to obtain R1 to R9:
[0040] S301: Estimated defocus radius R irec ;
[0041] S302: Utilizing the defocus radius R irec Construct a point spread function and use it to control the Wiener filter function on the image. i Perform defocus restoration to obtain the image. irec ;
[0042] S303: Use the Sobel operator on the image irec Perform edge detection, and then calculate the image. irec gradient mean G irec ;
[0043] S304: Calculate G irec With G a The squared error value, where the squared error value refers to the square of the difference; when the squared error value is less than the preset threshold A, R is used. irec As R i When the squared error value is not less than the preset threshold A, adjust the defocus radius R. irec Then proceed with S302-S304 again.
[0044] S4: Based on Dis1 to Dis9, and Dis a Calculate ΔDis respectively i =Dis a -Dis iWe obtain ΔDis1 to ΔDis9, and use R1 to R9 and ΔDis1 to ΔDis9 to obtain R through the least squares estimation algorithm. i With ΔDis i The mapping relationship is used as the first mapping relationship. S5: Replace the standard sample with the wafer to be photographed;
[0045] S6: The wafer is moved by a displacement stage, so that an area of the wafer to be photographed is located within the imaging area of the camera, and an image is captured using the camera. b Record the distance between the corresponding camera and the wafer in the Z direction. b .
[0046] S7: Calculate ΔDis b =Dis a -Dis b , will ΔDis b Input the first mapping relationship, and the first mapping relationship outputs the corresponding defocus radius R. b Using the defocus radius R b Construct a point spread function and use it to control the Wiener filter function on the image. b Perform defocus restoration to obtain the restored image. b .
[0047] Through the above process, the defocusing restoration of the wafer imaging image was completed.
[0048] Furthermore, if a first mapping relationship already exists, steps S5-S7 can be performed separately to further save steps and improve efficiency.
[0049] Additionally, S8 can be further configured: this involves S6-S7, using multiple images. b The images are pieced together to form a larger image of the wafer.
[0050] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Therefore, the above descriptions are merely embodiments of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the present invention. Various equivalent changes and modifications are included without departing from the spirit and scope of the present invention, and all such changes and modifications fall within the scope of the present invention as claimed.
Claims
1. A method for repairing defocus in wafer imaging, characterized in that, include: Move the standard sample in the XY plane and take n images of the standard sample at different positions to obtain the images. i And record the corresponding distance Dis between the camera and the standard in the Z direction. i Where i = 1, 2, 3...n; Find the clearest image a and the corresponding Dis a ; For each image i The sharpness of the restored out-of-focus image is calculated to be closest to that of the original image. a defocus radius R i Calculate ΔDis i =Dis a -Dis i Fitting R i With ΔDis i To obtain the first mapping relationship; Move the wafer to the preset position and take an image of the wafer. b And record the corresponding distance Dis b Calculate ΔDis b =Dis a -Dis b According to ΔDis b The first mapping relationship yields the corresponding defocus radius R. b Using R b For Image b The image is restored to its original state, resulting in a wafer imaging image after the defocused image has been repaired.
2. The wafer imaging defocusing repair method as described in claim 1, characterized in that: The standard sample has feature points set on it, and each image is a feature point set on it. i It contains at least 3 feature points.
3. The wafer imaging defocus repair method as described in claim 2, characterized in that: Each image i In the given information, there are at least three non-collinear feature points.
4. The wafer imaging defocus repair method as described in claim 1, characterized in that, The defocus radius R is obtained as follows: i Estimated defocus radius R irec Using R irec Perform out-of-focus image restoration to obtain the restored image. irec Comparison Image irec With Image a When the difference in sharpness between the two meets the preset conditions, the corresponding defocus radius R is used. irec As the defocus radius R i .
5. The wafer imaging defocusing repair method as described in claim 4, characterized in that, Compare Images in the following way irec With Image a Sharpness: for Image irec Image a Edge detection and the mean gradient of the image are calculated sequentially to obtain the Image. irec Image a The corresponding gradient mean G irec G a Calculate G irec With G a The squared error value is considered to meet the preset condition when the squared error value is less than the preset threshold.
6. The wafer imaging defocus repair method as described in claim 4, characterized in that: The estimated defocus radius R is obtained by traversing within a preset range with a fixed step size. irec .
7. A wafer imaging defocus repair method as described in any one of claims 1 or 4, characterized in that: Wiener filtering algorithm is used for out-of-focus image restoration.
8. The wafer imaging defocusing repair method as described in claim 1, characterized in that, The method further includes: moving the wafer to multiple preset positions and capturing multiple wafer images. b The wafer imaging images after defocus image restoration were calculated and obtained respectively.
9. The wafer imaging defocusing repair method as described in claim 1, characterized in that: In capturing images i At this time, the object being photographed forms an angle with the XY plane.
10. The wafer imaging defocusing repair method as described in claim 9, characterized in that: The angle between the photographed surface of the standard sample and the XY plane shall not be less than the angle between the photographed surface of the wafer and the XY plane.
Citation Information
Patent Citations
Method for repairing defocusing of imaging blocks on edge of wafer
CN102129174A
Out-of-focus two-dimensional code image restoration method based on edge prior
CN112200741A
