Wafer image preprocessing method, device and equipment and storage medium

Abstract

The application discloses a wafer image preprocessing method, device and equipment and a storage medium, and the method comprises the following steps: acquiring an initial inspection image of a wafer to be measured; performing noise distribution estimation on the initial inspection image, and positioning a highlight defect area in the initial inspection image; performing smoothing processing on the highlight defect area to obtain a smoothed image and a difference image of the highlight defect area and a corresponding area on the smoothed image; performing a frequency domain filtering operation on the smoothed image to obtain a filtered image; and superimposing the difference image on the corresponding area on the filtered image to obtain a final inspection image. According to the application, the highlight defect area in the wafer diffraction image is positioned and separated, then the frequency domain filtering operation is performed on the remaining part of the image, finally, the difference image corresponding to the highlight defect area is superimposed on the filtered image, the image reconstruction is completed, the morphological characteristics of the highlight defect are retained, and the subsequent defect detection and classification are not affected.

Description

Technical Field

[0001] This application relates to the field of wafer inspection technology, and in particular to a wafer image preprocessing method, apparatus, device and storage medium. Background Technology

[0002] When a wafer is illuminated by a laser spot, it produces reflected or scattered light. This light is collected by the imaging optical path and then shines onto the camera's photosensitive source, forming an image containing information about the wafer's background, structure, and defects. By detecting defects in this image, the wafer's yield can be determined. In addition to the background, structure, and defect information, the image typically contains noise introduced by the laser, wafer processing, and the camera. This noise can interfere with defect detection. Filtering out periodic noise is usually done in the frequency domain. However, the image after inverse transformation of the frequency domain image often exhibits diffraction fringes in bright areas, damaging the structural characteristics of defects and affecting subsequent defect classification processes.

[0003] Currently, the common approach to address image distortion in bright areas is to superimpose a Gaussian filter or other smoothing filter onto the filter. This method can reduce the distortion of the image after inverse transformation to some extent. However, the smoothing of the filter can affect the effect of frequency domain filtering to some extent and is not effective enough in reducing the distortion of bright defects, thus still having a significant impact on the accuracy of wafer defect detection. Summary of the Invention

[0004] In view of this, this application provides a wafer image preprocessing method, apparatus, device and storage medium to solve the problem of image distortion due to high brightness defects in existing wafer inspection image preprocessing.

[0005] To address the aforementioned technical problems, this application provides a wafer image preprocessing method, comprising: acquiring an initial image of the wafer to be inspected; estimating the noise distribution of the initial image to be inspected to locate the bright defect region in the initial image to be inspected; smoothing the bright defect region to obtain a smoothed image and a difference image between the bright defect region and the corresponding region on the smoothed image; performing a frequency domain filtering operation on the smoothed image to obtain a filtered image; and superimposing the difference image onto the corresponding region on the filtered image to obtain the final image to be inspected.

[0006] As a further improvement of this application, noise distribution estimation is performed on the initial image to be inspected to locate the bright defect region in the initial image to be inspected, including: extracting the reference pixel values ​​of all reference pixels within a preset neighborhood of each pixel, and calculating the first feature value and the second feature value of each pixel using the reference pixel values; calculating the pixel threshold of each pixel based on the first feature value and the second feature value; and taking the region formed by pixels whose pixel values ​​exceed the pixel threshold as the bright defect region.

[0007] As a further improvement of this application, the first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood, and the second feature value includes the first variance of all reference pixel values ​​within the preset neighborhood; the calculation process of the pixel threshold is expressed as follows:

[0008] T i =medain i +coeff1*std i ;

[0009] Among them, T i Medain represents the pixel threshold of the i-th pixel. i std represents the median or mean value corresponding to the i-th pixel. i represents the first variance value corresponding to the i-th pixel, and coeff1 represents the pre-defined noise estimation coefficients.

[0010] As a further improvement of this application, the first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood, and the second feature value includes the second variance value among all reference pixel values ​​within the preset neighborhood, the second variance value being expressed as:

[0011] std j =coeff2*(medain) j -min j );

[0012] Among them, std j Medain represents the second variance value corresponding to the j-th pixel. j min represents the median or mean value corresponding to the j-th pixel. j represents the minimum pixel value within the preset neighborhood range corresponding to the j-th pixel, and coeff2 represents the pre-calibrated parameters;

[0013] The calculation process for the pixel threshold is expressed as follows:

[0014] T j =medain j +coeff1*[coeff2*(medain j -min j )];

[0015] Among them, T j represents the pixel threshold of the j-th pixel, and coeff1 represents the pre-defined noise estimation coefficients.

[0016] As a further improvement of this application, the preset neighborhood range is represented as all pixels within a window of size M*N constructed with the pixel as the center.

[0017] As a further improvement of this application, the highlighted defect region is smoothed to obtain a smooth image and a difference image between the highlighted defect region and the corresponding region on the smooth image, including: obtaining the target pixel values ​​of all target pixels in the highlighted defect region; calculating the difference between the target pixel value corresponding to each target pixel and the corresponding first feature value, and constructing a difference image based on the difference; replacing the pixel value of each target pixel in the highlighted defect region with the corresponding first feature value to obtain the smooth image.

[0018] To address the aforementioned technical problems, another technical solution adopted in this application is: providing a wafer image preprocessing apparatus, comprising: an acquisition module for acquiring an initial image of the wafer to be inspected; a positioning module for estimating the noise distribution of the initial image to be inspected and locating the bright defect region in the initial image to be inspected; a smoothing module for smoothing the bright defect region to obtain a smoothed image and a difference image between the bright defect region and the corresponding region on the smoothed image; a filtering module for performing frequency domain filtering on the smoothed image to obtain a filtered image; and a reconstruction module for superimposing the difference image onto the corresponding region on the filtered image to obtain the final image to be inspected.

[0019] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide a computer device, the computer device including a processor and a memory coupled to the processor, the memory storing program instructions, and when the program instructions are executed by the processor, causing the processor to perform the steps of the wafer image preprocessing method as described above.

[0020] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a storage medium storing program instructions capable of implementing the wafer image preprocessing method described above.

[0021] The beneficial effects of this application are as follows: The wafer image preprocessing method of this application locates and separates the bright defect region in the wafer diffraction image, then performs frequency domain filtering on the rest of the image, and finally superimposes the difference image corresponding to the bright defect region onto the filtered image to complete the image reconstruction. The reconstructed image retains the morphological features of the bright defect while filtering out noise, and will not affect the subsequent defect detection and classification, thereby improving the accuracy of defect detection. Attached Figure Description

[0022] Figure 1 This is a schematic flowchart of a wafer image preprocessing method according to an embodiment of the present invention;

[0023] Figure 2 This is a schematic diagram of the functional modules of the wafer image preprocessing device according to an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of the structure of the storage medium according to an embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0027] The terms "first," "second," and "third" in this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0028] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0029] Figure 1 This is a schematic flowchart of a wafer image preprocessing method according to an embodiment of the present invention. It should be noted that if substantially the same result is obtained, the method of the present invention is not necessarily identical. Figure 1 The illustrated process sequence is limited. For example... Figure 1 As shown, the wafer image preprocessing method includes the following steps:

[0030] Step S1: Obtain the initial image of the wafer to be tested.

[0031] It is important to note that in the semiconductor manufacturing industry, wafer inspection is crucial for ensuring product quality. Wafer diffraction images can provide information about the microstructure and defects on the wafer surface. However, the initially detected diffraction images are often affected by factors such as noise, making direct defect detection based on the initial images too inaccurate. Therefore, preprocessing of the initial images is necessary to improve the accuracy of subsequent analysis.

[0032] Specifically, in practical applications, images of the wafer are acquired using specialized image acquisition equipment, such as a high-resolution microscope, to obtain an initial image to be inspected. This initial image includes, but is not limited to, information such as background structure, periodic noise (e.g., laser interference fringes), and bright defects (e.g., micron-sized scratches or particulate contamination).

[0033] Step S2: Estimate the noise distribution of the initial image to be inspected and locate the bright defect areas in the initial image to be inspected.

[0034] Specifically, when locating the highlighted defect region, the pixel value of each pixel in the initial image to be inspected is first obtained; then, noise distribution is estimated based on the pixel value, which can be achieved through Gaussian noise model, wavelet transform-based noise estimation model, etc.; next, a suitable threshold is set according to the noise distribution, which can be determined in various ways, such as based on empirical values, statistical analysis of known defect images, or by using an adaptive thresholding method (e.g., setting an adaptive threshold based on local mean and standard deviation, expressed as: T=μ+k*σ, where T represents the adaptive threshold, μ represents the local mean, σ represents the local standard deviation, and k represents the empirical coefficient); finally, based on the threshold, the entire image is traversed, and pixels from the pixel value to those exceeding the threshold are marked. All marked pixels constitute the highlighted defect region.

[0035] Furthermore, it should be understood that the present invention performs pre-inspection on the initial image to be inspected, and its goal is to locate the bright defect region. Therefore, the noise estimation accuracy of the initial image to be inspected does not need to be too high, as long as the bright defect region can be located. Therefore, step S2 specifically includes:

[0036] 1. Extract the reference pixel values ​​of all reference pixels within the preset neighborhood range of each pixel, and use the reference pixel values ​​to calculate the first feature value and the second feature value of each pixel.

[0037] 2. Calculate the pixel threshold for each pixel based on the first feature value and the second feature value.

[0038] 3. The area formed by pixels whose pixel values ​​exceed the pixel threshold is defined as the highlight defect area.

[0039] In some embodiments, to ensure the accuracy of image reconstruction, the preset neighborhood range is represented as all pixels within a window of size M*N centered at the pixel. Therefore, it is necessary to traverse each pixel in the initial image to be inspected to obtain the reference pixel values ​​of all reference pixels within a window of size M*N centered at each pixel. Here, the values ​​of M and N are both in the range [9, 121].

[0040] It should be noted that when traversing the pixels at the edge positions of the initial image to be inspected, the M*N window range exceeds the edge of the initial image to be inspected. Therefore, it is necessary to fill the pixels in the excess area. The filling methods can include the following:

[0041] 1. Mirror Fill: First, for each pixel to be processed, calculate the number of pixels to be filled in the horizontal and vertical directions based on the size of the M×N window. Then, starting from the image edge, copy the pixel values ​​sequentially in a mirror manner. For example, in the horizontal direction, if the current pixel is on the far right, then the mirror pixel to its right is the pixel value at its symmetrical position to its left; the same applies to the vertical direction. Finally, after obtaining the mirror pixels, fill the window area that extends beyond the image edge.

[0042] 2. Constant Fill. Constant fill fills areas that extend beyond the image boundary with a fixed constant value. This constant value can be selected based on the specific application scenario, such as the average pixel value of the image, 0, or 255.

[0043] 3. Edge copy fill. Similar to mirror fill, edge copy fill calculates the number of pixels that need to be filled in the horizontal and vertical directions, and then copies the pixel values ​​of the image edges to the parts that extend beyond the boundaries.

[0044] In this embodiment, by traversing all pixels and then locating the bright defect area based on the traversal results, the positioning accuracy is high, but the computational load is large.

[0045] Furthermore, to improve reconstruction efficiency, in some embodiments, when traversing pixels within a preset neighborhood of size M*N, the two-dimensional convolution can be decomposed into two one-dimensional convolutions. First, a window of size M*1 is used to traverse the pixels within the preset neighborhood, and then a window of size 1*N is used to traverse the pixels within the preset neighborhood. The computational complexity of the original M*N two-dimensional convolution is O = (M×N×H×W), which becomes O = ((M+N)×H×W) after decomposition. As the preset neighborhood becomes larger, the computational complexity after decomposition is significantly lower than that before decomposition, thus improving computational efficiency. Therefore, when the preset neighborhood is large, it is preferable to use this method of decomposing the two-dimensional convolution into one-dimensional convolutions, sacrificing a small amount of precision for a significant improvement in computational efficiency.

[0046] It should be noted that in this embodiment, the areas exceeding the limit also need to be filled with pixels. Please refer to the above embodiments for the filling method, which will not be repeated here.

[0047] Furthermore, based on the above embodiments, in some embodiments, the first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood range, and the second feature value includes the first variance of all reference pixel values ​​within the preset neighborhood range.

[0048] The calculation process for the pixel threshold is expressed as follows:

[0049] T i =medain i +coeff1*std i ;

[0050] Among them, T i Medain represents the pixel threshold of the i-th pixel. i std represents the median or mean value corresponding to the i-th pixel. i Let represent the first variance value corresponding to the i-th pixel, and coeff1 represent the pre-defined noise estimation coefficient. It's important to understand that this noise estimation coefficient controls the weight of the standard deviation in the threshold calculation. When coeff1 is low, it means the standard deviation has a smaller impact on the threshold. In this case, the threshold relies more on the median, and the segmentation result is more sensitive to defects because a smaller coeff1 results in a relatively low threshold, making it easier to identify some small differences that might otherwise be ignored as defects. However, at the same time, the lower threshold also increases the possibility of misclassifying noise as defects. When coeff1 is high, the standard deviation plays a larger role in the threshold calculation, and the threshold will be correspondingly higher. This makes it less sensitive to defect segmentation but effectively reduces the possibility of noise being misclassified as defects.

[0051] Furthermore, based on the above embodiments, in some other embodiments, the first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood, and the second feature value includes the second variance value among all reference pixel values ​​within the preset neighborhood, wherein the second variance value is expressed as:

[0052] std j =coeff2*(medain) j -min j );

[0053] Among them, std j Medain represents the second variance value corresponding to the j-th pixel. j min represents the median or mean value corresponding to the j-th pixel. j represents the minimum pixel value within the preset neighborhood range corresponding to the j-th pixel, and coeff2 represents the pre-calibrated parameters;

[0054] The calculation process for the pixel threshold is expressed as follows:

[0055] T j =medain j +coeff1*[coeff2*(medain j -min j )];

[0056] Among them, T j represents the pixel threshold of the j-th pixel, and coeff1 represents the pre-defined noise estimation coefficients.

[0057] Step S3: Smooth the highlighted defect area to obtain a smoothed image and a difference image between the highlighted defect area and the corresponding area on the smoothed image.

[0058] Specifically, in order to reduce the distortion problem of bright defect areas in the image under inspection by frequency domain filtering operation, this embodiment performs smoothing processing on all pixels in the bright defect areas, thereby suppressing the brightness of the bright defect areas while preserving the pixel difference between the bright defect areas and the smoothed image.

[0059] Furthermore, step S3 specifically includes:

[0060] 1. Obtain the target pixel values ​​of all target pixels within the highlighted defect area.

[0061] Specifically, each pixel within the highlighted defect area is traversed, and its pixel value is recorded as the target pixel value.

[0062] 2. Calculate the difference between the target pixel value and the corresponding first feature value for each target pixel, and construct a difference image based on the difference.

[0063] Specifically, for each target pixel in the bright defect area, the first feature value corresponding to the pixel (such as the median or mean value calculated in the above embodiment) is subtracted from its target pixel value to obtain the difference. A difference image is constructed based on these differences. The value of each pixel in the difference image reflects the degree of deviation of the point from the smoothing reference value.

[0064] 3. Replace the pixel value of each target pixel in the highlighted defect area with the corresponding first feature value to obtain a smooth image.

[0065] Specifically, by replacing the pixel value of each target pixel in the bright defect area with the corresponding first feature value, the bright defect area becomes smoother, avoiding distortion problems during subsequent frequency domain filtering operations.

[0066] Step S4: Perform frequency domain filtering on the smoothed image to obtain the filtered image.

[0067] Specifically, frequency domain filtering is an effective image processing method that can remove noise or unwanted periodic patterns from specific frequency components of an image. The smoothed image is converted to the frequency domain, for example, using methods such as the Discrete Fourier Transform (DFT). In the frequency domain, appropriate filters are designed based on the characteristics of the noise or interference, such as low-pass filters (for removing high-frequency noise), high-pass filters (for enhancing high-frequency information such as edges), or band-pass filters (allowing frequency components in specific bands to pass through). The frequency domain image data after filtering is then subjected to operations such as inverse Fourier transform to obtain a filtered image that retains useful information while reducing noise and interference.

[0068] Step S5: Overlay the difference image onto the corresponding region of the filtered image to obtain the final image to be inspected.

[0069] Specifically, based on the positional correspondence between the difference image and the highlighted defect area, the difference image is accurately superimposed onto the corresponding highlighted defect area on the filtered image. Through pixel-by-pixel superposition, combining the advantages of smoothing and frequency domain filtering, the image's feature information is preserved while highlighting defect information, ultimately yielding the final image to be inspected. This final image to be inspected provides a more accurate data foundation for subsequent defect analysis, dimensional measurement, and other processing.

[0070] The wafer image preprocessing method in this embodiment locates and separates the bright defect regions in the wafer diffraction image, then performs frequency domain filtering on the remaining part of the image, and finally superimposes the difference image corresponding to the bright defect region onto the filtered image to complete image reconstruction. The reconstructed image retains the morphological features of the bright defects while filtering out noise, and will not affect the subsequent detection and classification of defects, thus improving the accuracy of defect detection.

[0071] Figure 2 This is a schematic diagram of the functional modules of the wafer image preprocessing apparatus according to an embodiment of the present invention. Figure 2 As shown, the wafer image preprocessing device 20 includes: an acquisition module 21, a positioning module 22, a smoothing module 23, a filtering module 24, and a reconstruction module 25.

[0072] The acquisition module 21 is used to acquire the initial image of the wafer to be inspected;

[0073] The positioning module 22 is used to estimate the noise distribution of the initial image to be inspected and locate the bright defect area in the initial image to be inspected.

[0074] Smoothing module 23 is used to smooth the bright defect area to obtain a smooth image and a difference image between the bright defect area and the corresponding area on the smooth image.

[0075] Filtering module 24 is used to perform frequency domain filtering on the smoothed image to obtain a filtered image;

[0076] The reconstruction module 25 is used to overlay the difference image onto the corresponding region of the filtered image to obtain the final image to be inspected.

[0077] Optionally, the localization module 22 performs noise distribution estimation on the initial image to be inspected and locates the bright defect region in the initial image to be inspected. Specifically, it includes: extracting the reference pixel values ​​of all reference pixels within a preset neighborhood of each pixel, and calculating the first feature value and the second feature value of each pixel using the reference pixel values; calculating the pixel threshold of each pixel based on the first feature value and the second feature value; and taking the region formed by pixels whose pixel values ​​exceed the pixel threshold as the bright defect region.

[0078] Optionally, the first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood, and the second feature value includes the first variance of all reference pixel values ​​within the preset neighborhood; the calculation process of the pixel threshold is expressed as follows:

[0079] T i =medain i +coeff1*std i ;

[0080] Among them, T i Medain represents the pixel threshold of the i-th pixel. i std represents the median or mean value corresponding to the i-th pixel. i represents the first variance value corresponding to the i-th pixel, and coeff1 represents the pre-defined noise estimation coefficients.

[0081] Optionally, the first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood, and the second feature value includes the second variance of all reference pixel values ​​within the preset neighborhood, wherein the second variance is expressed as:

[0082] std j =coeff2*(medain) j -min j );

[0083] Among them, std j Medain represents the second variance value corresponding to the j-th pixel. j min represents the median or mean value corresponding to the j-th pixel. j represents the minimum pixel value within the preset neighborhood range corresponding to the j-th pixel, and coeff2 represents the pre-calibrated parameters;

[0084] The calculation process for the pixel threshold is expressed as follows:

[0085] T j =medain j +coeff1*[coeff2*(medain j -min j )];

[0086] Among them, T j represents the pixel threshold of the j-th pixel, and coeff1 represents the pre-defined noise estimation coefficients.

[0087] Optionally, the preset neighborhood range is represented as all pixels within a window of size M*N constructed with the pixel as the center.

[0088] Optionally, the smoothing module 23 performs smoothing processing on the bright defect area to obtain a smooth image and a difference image between the bright defect area and the corresponding area on the smooth image. Specifically, this includes: obtaining the target pixel values ​​of all target pixels in the bright defect area; calculating the difference between the target pixel value corresponding to each target pixel and the corresponding first feature value, and constructing a difference image based on the difference; and replacing the pixel value of each target pixel in the bright defect area with the corresponding first feature value to obtain the smooth image.

[0089] For further details regarding the implementation techniques of each module in the wafer image preprocessing apparatus of the above embodiments, please refer to the description in the wafer image preprocessing method of the above embodiments, which will not be repeated here.

[0090] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For apparatus embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0091] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. Figure 3 As shown, the computer device 30 includes a processor 31 and a memory 32 coupled to the processor 31. The memory 32 stores program instructions. When the program instructions are executed by the processor 31, the processor 31 performs the wafer image preprocessing method steps described in any of the above embodiments.

[0092] The processor 31 can also be referred to as a Central Processing Unit (CPU). The processor 31 may be an integrated circuit chip with signal processing capabilities. The processor 31 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor.

[0093] See Figure 4 , Figure 4 This is a schematic diagram of the structure of a storage medium according to an embodiment of the present invention. The storage medium of this embodiment stores program instructions 41 capable of implementing the above-described wafer image preprocessing method. These program instructions 41 can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or computer devices such as computers, servers, mobile phones, and tablets.

[0094] In the several embodiments provided in this application, it should be understood that the disclosed computer devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0095] Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A wafer image preprocessing method, characterized in that, It includes: Obtain the initial image of the wafer to be inspected; The initial image to be inspected is subjected to noise distribution estimation. The pixel threshold of each pixel is calculated based on the reference pixel value within a preset neighborhood range. The region formed by pixels whose pixel values ​​exceed the pixel threshold is taken as the bright defect region in the initial image to be inspected. The highlighted defect area is smoothed to obtain a smoothed image and a difference image between the highlighted defect area and the corresponding area on the smoothed image. The smoothed image is subjected to frequency domain filtering to obtain a filtered image; The difference image is superimposed onto the corresponding region of the filtered image to obtain the final image to be inspected.

2. The wafer image preprocessing method according to claim 1, characterized in that, The step of calculating the pixel threshold based on reference pixel values ​​within a preset neighborhood range for each pixel includes: Extract the reference pixel values ​​of all reference pixels within a preset neighborhood range for each pixel; The first feature value and the second feature value of each pixel are calculated using the reference pixel value; The pixel threshold for each pixel is calculated based on the first feature value and the second feature value.

3. The wafer image preprocessing method according to claim 2, characterized in that, The first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood range, and the second feature value includes the first variance of all reference pixel values ​​within a preset neighborhood range. The calculation process for the pixel threshold is expressed as follows: ； in, Indicates the first Pixel threshold for each pixel Indicates the first The median or mean value corresponding to each pixel. Indicates the first The first variance value corresponding to each pixel. The table contains pre-defined noise estimation coefficients.

4. The wafer image preprocessing method according to claim 2, characterized in that, The first feature value includes the median or mean of all reference pixel values ​​within a preset neighborhood range, and the second feature value includes the second variance of all reference pixel values ​​within the preset neighborhood range, wherein the second variance is expressed as: ； in, Indicates the first The second variance value corresponding to each pixel. Indicates the first The median or mean value corresponding to each pixel. Indicates the first The minimum pixel value within a preset neighborhood range corresponding to each pixel. Indicates pre-calibrated parameters; The calculation process for the pixel threshold is expressed as follows: [ ]； in, Indicates the first Pixel threshold for each pixel The table contains pre-defined noise estimation coefficients.

5. The wafer image preprocessing method according to claim 4, characterized in that, The preset neighborhood range is defined as all pixels within a window of size M*N, centered on a single pixel.

6. The wafer image preprocessing method according to claim 5, characterized in that, The step of smoothing the highlighted defect region to obtain a smoothed image and a difference image between the highlighted defect region and the corresponding region on the smoothed image includes: Obtain the target pixel values ​​of all target pixels within the highlighted defect area; Calculate the difference between the target pixel value and the corresponding first feature value for each target pixel, and construct the difference image based on the difference; The pixel value of each target pixel in the highlighted defect area is replaced with the corresponding first feature value to obtain the smoothed image.

7. A wafer image preprocessing apparatus, characterized in that, It includes: The acquisition module is used to acquire the initial image of the wafer under test. The positioning module is used to estimate the noise distribution of the initial image to be inspected, calculate the pixel threshold of each pixel based on the reference pixel value within a preset neighborhood range, and take the region formed by pixels whose pixel values ​​exceed the pixel threshold as the bright defect region in the initial image to be inspected. A smoothing module is used to smooth the bright defect area to obtain a smoothed image and a difference image between the bright defect area and the corresponding area on the smoothed image. The filtering module is used to perform frequency domain filtering on the smoothed image to obtain a filtered image; The reconstruction module is used to overlay the difference image onto the corresponding region of the filtered image to obtain the final image to be inspected.

8. A computer device, characterized in that, The computer device includes a processor and a memory coupled to the processor, the memory storing program instructions that, when executed by the processor, cause the processor to perform the steps of the wafer image preprocessing method as described in any one of claims 1-6.

9. A storage medium, characterized in that, The device stores program instructions capable of implementing the wafer image preprocessing method as described in any one of claims 1-6.

Images