Image processing method and apparatus

Abstract

Embodiments of the present application provide an image processing method, device, computer equipment and computer readable storage medium, and belong to the field of semiconductor manufacturing. The image processing method comprises: acquiring an initial image of a target silicon wafer, the initial image being spliced from a plurality of field of view images, one field of view image being obtained by one-time scanning; performing low-pass filtering processing on the initial image to obtain a light intensity distribution image corresponding to the initial image, the light intensity distribution image being used to represent uneven light intensity distribution on the initial image; and processing the initial image according to the light intensity distribution image to remove splicing traces of the plurality of field of view images, to obtain a target image corresponding to the initial image. The technical scheme of the embodiments of the present application can reduce the hardware cost of image processing and improve the flexibility of image processing.

Description

Technical Field

[0001] This application relates to the field of semiconductor manufacturing technology, and in particular to an image processing method, apparatus, computer equipment, and computer-readable storage medium. Background Technology

[0002] In the semiconductor manufacturing industry, quality inspection of the back side of silicon wafers has a significant impact on chip yield. To achieve complete digital inspection of the back side of silicon wafers, image stitching technology can be used to stitch together multiple field-of-view images acquired through multiple scans into a complete global image, thus meeting the dual requirements of high resolution and a large field of view.

[0003] However, due to differences in lighting conditions at different scanning locations, the stitched images often exhibit noticeable stitching marks, affecting the accuracy of subsequent defect detection. To eliminate these stitching marks, illumination homogenization can be achieved by optimizing the imaging optical system (such as the camera, lens, and light source). While stable and reliable, this method is costly and lacks flexibility. Alternatively, a reference image can be obtained using a constant grayscale correction plate for algorithmic correction, but the correction effect significantly decreases when the correction plate is noisy or damaged.

[0004] It should be noted that the above content is not necessarily prior art, nor is it intended to limit the scope of patent protection of this application. Summary of the Invention

[0005] This application provides an image processing method, apparatus, computer device, and computer-readable storage medium to solve or alleviate one or more of the technical problems mentioned above.

[0006] One aspect of this application provides an image processing method, the method comprising: An initial image of the target silicon wafer is acquired, the initial image being composed of multiple field-of-view images stitched together, and one of the field-of-view images is obtained through a single scan; The initial image is subjected to low-pass filtering to obtain a light intensity distribution image corresponding to the initial image, and the light intensity distribution image is used to characterize the non-uniform light intensity distribution on the initial image. The initial image is processed based on the light intensity distribution image to remove stitching marks from the multiple field-of-view images, thereby obtaining the target image corresponding to the initial image.

[0007] Optionally, an initial image of the target silicon wafer is acquired, including: The number of circular scans and the path length of each circular scan are determined based on the diameter of the target silicon wafer and the width of the camera's field of view. Multiple scan coordinates are determined based on the path length of each circular scan, the length of the camera's field of view, and the preset overlap range; The target silicon wafer is scanned by a camera according to the multiple scanning coordinates to obtain the multiple field-of-view images; The multiple field-of-view images are stitched together according to the preset overlap range to obtain the initial image.

[0008] Optionally, the initial image is subjected to low-pass filtering, including: Based on the imaging quality of the initial image, determine the Gaussian kernel size parameter; Based on the Gaussian kernel size parameters, obtain the target Gaussian kernel function; The initial image is low-pass filtered using the target Gaussian kernel function; The target Gaussian kernel function is used to filter high-frequency texture information in the initial image and retain low-frequency light intensity information in the initial image.

[0009] Optionally, the initial image is subjected to low-pass filtering using the target Gaussian kernel function, including: Obtain the target grayscale image corresponding to the initial image; Perform a Fourier transform on the target grayscale image to obtain its frequency domain representation; The frequency domain representation of the light intensity distribution image is obtained based on the frequency domain representation of the target grayscale image and the target Gaussian kernel function; The light intensity distribution image is obtained by performing an inverse Fourier transform on its frequency domain representation.

[0010] Optionally, obtaining the target grayscale image corresponding to the initial image includes: Obtain the initial grayscale image corresponding to the initial image; The boundaries of the initial grayscale image are extended by pixels to obtain the target grayscale image.

[0011] Optionally, the initial image is processed based on the light intensity distribution image, including: Divide the initial pixel value at each pixel position in the initial image by the light intensity pixel value at the corresponding pixel position in the light intensity distribution image to obtain the normalized pixel value. The target pixel value at the corresponding pixel position in the target image is obtained based on the normalized pixel value and the preset scaling factor.

[0012] Optionally, the method further includes: The target image is subjected to grayscale linear processing to adjust the grayscale range of the target image to a preset grayscale range.

[0013] Another aspect of this application provides an image processing apparatus, the apparatus comprising: An acquisition module is used to acquire an initial image of the target silicon wafer, wherein the initial image is composed of multiple field-of-view images stitched together, and one field-of-view image is obtained through a single scan; The processing module is used to perform low-pass filtering on the initial image to obtain a light intensity distribution image corresponding to the initial image, wherein the light intensity distribution image is used to characterize the non-uniform light intensity distribution on the initial image; The removal module is used to process the initial image based on the light intensity distribution image to remove the stitching marks of the multiple field images and obtain the target image corresponding to the initial image.

[0014] Another aspect of this application provides a computer device, including: At least one processor; and A memory that is communicatively connected to the at least one processor; Wherein: the memory stores instructions that can be executed by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0015] Another aspect of this application provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the method described above.

[0016] The embodiments of this application employing the above technical solution may include the following advantages: low-pass filtering is applied to the initial image to obtain a light intensity distribution map, and the splicing marks on the initial image are removed using the light intensity distribution map to obtain the target image. Therefore, by directly applying low-pass filtering to the initial image to obtain its non-uniform light intensity distribution, the dependence on a constant grayscale correction plate can be reduced, lowering the hardware cost of image processing and improving the flexibility of image processing. Attached Figure Description

[0017] The accompanying drawings exemplify embodiments and form part of the specification, serving together with the textual description to explain exemplary implementations of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0018] Figure 1 This diagram schematically illustrates the operating environment of the image processing method according to Embodiment 1 of this application; Figure 2 A flowchart illustrating an image processing method according to Embodiment 1 of this application is shown schematically; Figure 3 Schematic illustration Figure 2 Flowchart of the sub-steps in step S200; Figure 4 This illustration schematically shows an operation diagram of scanning the back side of a target silicon wafer using a CCD camera; Figure 5 Schematic illustration Figure 2 Flowchart of the sub-steps in step S202; Figure 6 Schematic illustration Figure 5 Flowchart of the sub-steps in step S504; Figure 7(a) schematically illustrates an exemplary initial image; Figure 7(b) schematically shows an exemplary light intensity distribution image obtained under the condition of σ=13; Figure 7(c) schematically shows an exemplary light intensity distribution image obtained when σ=40; Figure 7(d) schematically shows an exemplary light intensity distribution image obtained when σ=100; Figure 8 Schematic illustration Figure 6 Flowchart of the sub-steps in step S600; Figure 9(a) schematically shows a dark-field grayscale image after extension and expansion; Figure 9(b) schematically shows a bright-field grayscale image after extension and expansion; Figure 9(c) schematically shows another dark-field grayscale image after extension and expansion; Figure 10(a) schematically shows an initial image of an exemplary target silicon wafer; Figure 10(b) schematically shows a target image corresponding to the initial image in Figure 10(a); Figure 10(c) schematically shows an initial image of another exemplary target silicon wafer; Figure 10(d) schematically shows a target image corresponding to the initial image in Figure 10(c); Figure 11 An exemplary application flowchart of the image processing method according to Embodiment 1 of this application is illustrated schematically; Figure 12 A block diagram of an image processing apparatus according to Embodiment 2 of this application is schematically shown; and Figure 13 A schematic diagram of the hardware architecture of a computer device according to Embodiment 3 of this application is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0020] It should be noted that the descriptions involving "first," "second," etc., in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0021] It should be noted that, in any stage of this application involving the collection, storage, use, transmission, and processing of data, each stage strictly adheres to the laws, regulations, industry standards, and regulatory requirements of the data source, usage location, and relevant countries and regions to ensure the legality and compliance of data activities. In the collection stage, the purpose, method, and scope of collection are clearly communicated to the data subject in a prominent manner. Collection is conducted only after obtaining the data subject's legal authorization, ensuring that the collection process follows the "minimum necessary" principle and does not exceed the scope of data collection. In the storage stage, storage periods are limited, and data is promptly deleted or anonymized / encrypted after the storage purpose is achieved. In the usage stage, a strict data security protection mechanism is implemented, using field-level desensitization technology and processing the original data according to preset desensitization rules. For different types of data, multiple desensitization strategies, such as data generalization, data anonymization, and data encryption, are employed to effectively mitigate the risk of sensitive information leakage and ensure that all data used is securely processed and desensitized, comprehensively protecting the rights and interests of data subjects and data security. In the transmission and processing stages, the confidentiality and security of data are ensured during transmission and processing.

[0022] In the description of this application, it should be understood that the numerical labels before the steps do not indicate the order of the steps, but are only used to facilitate the description of this application and to distinguish each step, and therefore should not be construed as a limitation of this application.

[0023] First, a definition of the terminology used in this application is provided: Silicon wafer: A thin, circular sheet made of high-purity single-crystal silicon.

[0024] Field of view image: refers to the image of the target area that the camera can capture during a single scan.

[0025] CCD (Charge-Coupled Device) camera: a high-sensitivity image sensor based on semiconductor technology.

[0026] Low-pass filtering: an image processing method that preserves low-frequency information and filters out high-frequency information.

[0027] Gaussian kernel: A filter generated based on a Gaussian function.

[0028] Gaussian kernel size: Controls the width of the Gaussian kernel and determines the degree of filtering. The larger the Gaussian kernel size, the stronger the filtering effect and the more blurred the image.

[0029] Fourier transform: A mathematical transformation that converts an image from the spatial domain (pixel coordinates) to the frequency domain.

[0030] Frequency domain: refers to the frequency representation of an image after Fourier transform, describing the distribution of different frequency components in the image.

[0031] Spatial domain: The pixel coordinate space of the original image.

[0032] Inverse Fourier transform: A mathematical transformation that restores an image from the frequency domain to the spatial domain.

[0033] Dark Field: Optical imaging mode, tilted illumination, imaging through scattered light, highly sensitive to defects such as tiny scratches and particles.

[0034] Bright field: Optical imaging mode, vertical illumination, suitable for observing surface defects.

[0035] Secondly, to facilitate understanding of the technical solutions provided in the embodiments of this application by those skilled in the art, the relevant technologies are described below: In the semiconductor manufacturing industry, wafer quality directly determines the yield and reliability of the final product. Among these factors, the quality of the back side of the silicon wafer (including contaminants, micro-scratches, chipping, etc.) directly impacts chip performance and yield. Back-side image stitching technology aims to resolve the conflict between high resolution and a large field of view, thereby achieving complete and accurate digitization of the back side of the silicon wafer, providing a foundation for subsequent quality assessment and process control. One challenge in back-side image stitching technology lies in correcting image stitching marks.

[0036] Hardware-level correction of stitching artifacts involves optimizing imaging optics, including cameras, lenses, and light sources. While stable and reliable, this approach is relatively expensive. Algorithm-level correction, on the other hand, offers flexibility, intelligence, and cost-effectiveness, including flat-field correction and fade-in / fade-out blending. However, flat-field correction relies heavily on the quality of the corrected image; noisy images are less effective at correcting stitching artifacts. Fade-in / fade-out blending, often using weighted averaging in a few fixed directions, is less effective at correcting stitching artifacts in arbitrary directions caused by rotating the silicon wafer backside during stitching.

[0037] Therefore, this application provides an image processing technical solution. In this technical solution, (1) by directly performing low-pass filtering on the initial image to obtain the non-uniform light intensity distribution on it, the dependence on the constant grayscale correction plate can be reduced, the hardware cost of image processing can be reduced, and the flexibility of image processing can be improved; (2) by introducing a preset overlap range for scanning path planning, the data acquisition integrity and accuracy of the initial image can be improved; (3) by using the target Gaussian function to filter high-frequency textures and retain low-frequency information, while effectively eliminating the interference of the physical characteristics of the target silicon wafer itself, the non-uniform illumination component that causes splicing marks can be extracted quickly and accurately, improving the processing efficiency and accuracy of image processing; (4) by using the light intensity distribution image for pixel-level image correction, the splicing marks caused by rotational scanning splicing in the initial image can be accurately eliminated, thereby obtaining the target image that characterizes the real physical state of the target silicon wafer, improving the accuracy and quality of image processing. At the same time, by combining frequency domain Gaussian low-pass filtering with multiplicative illumination model image division correction, the limitation on the direction of splicing marks during image correction is avoided, improving the adaptability and flexibility of image processing. See below for details.

[0038] Finally, for ease of understanding, an exemplary operating environment is provided below.

[0039] like Figure 1 As shown, the operating environment diagram includes: computer equipment 10000.

[0040] Computer devices 10000 include various types of electronic devices, such as portable handheld devices, general-purpose computers (such as personal computers and laptops), workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, thin clients, various messaging devices, sensors, or other electronic devices. These computer devices can run various types and versions of software applications and operating systems, such as Microsoft Windows, Apple iOS, UNIX-like operating systems, Linux or Linux-like operating systems (such as Google Chrome OS); or include various mobile operating systems, such as Microsoft Windows, Mobile OS, iOS, Windows Phone, and Android. Portable handheld devices can include cellular phones, smartphones, tablets, personal digital assistants, etc. Wearable devices can include head-mounted displays (such as smart glasses), etc.

[0041] Computer device 10000 may include input / output interfaces. Input interfaces may include a touchpad, touchscreen, mouse, keyboard, image sensor, or other sensing elements. Input interfaces can be configured to receive user commands that cause computer device 10000 to perform various operations, such as scanning a target silicon wafer or processing an initial image. Output interfaces are used to output information to the user, such as displaying a target image.

[0042] The technical solutions of this application will be described below using computer device 10000 as the implementation subject through multiple embodiments. It should be understood that these embodiments can be implemented in many different forms and should not be construed as being limited to the embodiments described herein.

[0043] Example 1 Figure 2 A flowchart illustrating an image processing method according to Embodiment 1 of this application is shown schematically.

[0044] like Figure 2 As shown, the image processing method may include steps S200~S204, wherein: Step S200: Obtain an initial image of the target silicon wafer. The initial image is composed of multiple field-of-view images stitched together, and one field-of-view image is obtained through a single scan.

[0045] Step S202: Perform low-pass filtering on the initial image to obtain a light intensity distribution image corresponding to the initial image. The light intensity distribution image is used to characterize the non-uniform light intensity distribution on the initial image.

[0046] Step S204: Process the initial image according to the light intensity distribution image to remove the stitching marks of the multiple field images and obtain the target image corresponding to the initial image.

[0047] The image processing method provided in this embodiment performs low-pass filtering on the initial image to obtain a light intensity distribution map, and then uses the light intensity distribution map to remove stitching marks on the initial image to obtain the target image. Therefore, by directly performing low-pass filtering on the initial image to obtain its non-uniform light intensity distribution, the reliance on a constant grayscale correction plate can be reduced, lowering the hardware cost of image processing and improving the flexibility of image processing.

[0048] The following combination Figure 2 The steps in steps S200 to S204, as well as other optional steps, are described in detail.

[0049] Step S200 An initial image of the target silicon wafer is obtained, which is composed of multiple field-of-view images stitched together, and one field-of-view image is obtained through a single scan.

[0050] The initial image can be an image of the back side (non-circuit side) of the target silicon wafer, which may show contaminants, micro-scratches, chips, etc. on the target silicon wafer. In some embodiments, the target silicon wafer can be scanned using an area scan camera (such as a CCD camera) in conjunction with an optical lens to obtain a field-of-view image. Before scanning, the scanning position and other information for each scan can be determined based on data such as the camera's field-of-view parameters and the size of the target silicon wafer.

[0051] In this embodiment, multiple field-of-view images are acquired through multiple scans and then stitched together to obtain an initial image. This satisfies the high-resolution and large field-of-view imaging requirements when inspecting the target silicon wafer, ensuring that a complete and accurate initial image covering the entire back of the target silicon wafer can be obtained, thus improving the image quality and data integrity of the initial image.

[0052] As mentioned earlier, the scanning position for each scan can be calculated based on parameters such as the camera's field of view. An exemplary calculation method is provided below.

[0053] In optional embodiments, such as Figure 3 As shown, step S200 includes: S300, based on the diameter of the target silicon wafer and the width of the camera's field of view, determine the number of circular scans and the path length of each circular scan.

[0054] S302 determines multiple scanning coordinates based on the path length of each circular scan, the length of the camera's field of view, and the preset overlap range.

[0055] S304, the target silicon wafer is scanned by a camera according to the multiple scanning coordinates to obtain the multiple field-of-view images.

[0056] S306, the multiple field-of-view images are stitched together according to the preset overlap range to obtain the initial image.

[0057] In some embodiments, there may be partial overlap between two adjacent scanning paths. When scanning is performed using a camera or the like, the camera's rotation angle (tangential angle of the camera relative to the center of the target silicon wafer) can be determined based on the path length of the circular scan. In practice, the camera can be driven to move along the calculated circular scanning path using relevant equipment. Figure 4 An exemplary schematic diagram of scanning the back of a target silicon wafer using a CCD camera is shown. As illustrated, during the scanning process, the light source can be switched synchronously via the DF (dark field) and BF (bright field) interfaces of the optical system to obtain field images with different characteristics.

[0058] In this embodiment, scanning coordinates are calculated by combining the physical dimensions of the target silicon wafer with camera field-of-view parameters to plan the scanning path, thus achieving efficient and automated acquisition of images of large-size silicon wafers. Simultaneously, by introducing a preset overlap range for scanning path planning, the integrity and accuracy of the initial image data acquisition can be improved.

[0059] Step S202 The initial image is subjected to low-pass filtering to obtain a light intensity distribution image corresponding to the initial image. The light intensity distribution image is used to characterize the non-uniform light intensity distribution on the initial image.

[0060] Before performing low-pass filtering on the initial image, preprocessing operations such as noise estimation and suppression (e.g., adaptive median filtering based on local statistics, non-local mean denoising) can be performed on the initial image. In some embodiments, low-pass filtering can be implemented using Gaussian low-pass filtering, ideal low-pass filtering, or spatial large-scale mean / median filtering. In other embodiments, initial images of the same target silicon wafer at different exposure times can be acquired, and the light intensity distributions can be obtained separately and then weighted and fused to obtain the final light intensity distribution image.

[0061] In this embodiment, a light intensity distribution image is obtained by directly low-pass filtering the initial image, thus enabling the direct estimation of the non-uniform illumination distribution from the initial image itself. Utilizing the characteristic that the non-uniform illumination distribution corresponding to the stitching marks is low-frequency information, low-pass filtering extracts the low-frequency components reflecting the overall illumination non-uniformity, while filtering out interference from high-frequency details such as the surface texture of the target silicon wafer. Therefore, an accurate light intensity distribution image can be obtained without relying on a constant grayscale correction plate, improving the environmental adaptability and operational flexibility of image processing.

[0062] As mentioned earlier, low-pass filtering can be applied to the initial image in various ways. The following provides an exemplary low-pass filtering method.

[0063] In optional embodiments, such as Figure 5 As shown, step S202 includes: S500, determine the Gaussian kernel size parameter based on the imaging quality of the initial image.

[0064] S502, Obtain the target Gaussian kernel function based on the Gaussian kernel size parameter.

[0065] S504, the initial image is low-pass filtered using the target Gaussian kernel function. The target Gaussian kernel function filters out high-frequency texture information in the initial image while retaining low-frequency light intensity information.

[0066] The Gaussian kernel size parameter can be the size of the Gaussian kernel (e.g., 13, 15, 17, etc.). In some embodiments, the Gaussian kernel size parameter can be determined based on factors such as the graininess of the initial image, the background noise level, or the contrast of the stitching artifacts. For example, if the foreground texture graininess in the initial image is high, a larger Gaussian kernel size parameter can be set; if the contrast of the stitching artifacts in the initial image is low, a smaller Gaussian kernel size parameter can be set. In other embodiments, the Gaussian kernel size parameter can also be determined comprehensively based on factors such as camera resolution and field of view parameters, lens distortion, and light source type. After determining the Gaussian kernel size parameter, the target Gaussian kernel function can be obtained using the following formula. :

[0067] in, Let σ be the coordinates in the frequency domain, and σ be the Gaussian kernel size parameter. High-frequency texture information can be details in the initial image where grayscale values ​​change drastically, such as physical defects on the target silicon wafer like micro-scratches, contaminants, and chipping, as well as the surface texture or graininess of the target silicon wafer itself. Low-frequency light intensity information can be background areas in the initial image where grayscale values ​​change gently, such as uneven light intensity distribution caused by non-uniformity of the optical system, light source attenuation, or camera imaging characteristics, as well as stitching marks or color difference bands caused by rotational scanning stitching.

[0068] In this embodiment, the Gaussian kernel size parameter is dynamically determined based on the imaging quality of the initial image, enabling precise control and adaptive adjustment of the non-uniform light intensity distribution estimation process. This ensures that high-quality light intensity distribution images can be acquired for different imaging environments, improving the accuracy of image processing. Simultaneously, by using a target Gaussian function to filter high-frequency textures and retain low-frequency information, interference from the physical characteristics of the target silicon wafer itself is effectively eliminated, while the non-uniform illumination components causing stitching marks can be quickly and accurately extracted, improving the processing efficiency and accuracy of image processing.

[0069] In practical processing, low-pass filtering can be implemented using the target Gaussian kernel function in various ways. An exemplary implementation is provided below.

[0070] In optional embodiments, such as Figure 6 As shown, step S504 includes: S600, Obtain the target grayscale image corresponding to the initial image.

[0071] S602, Perform a Fourier transform on the target grayscale image to obtain the frequency domain representation of the target grayscale image.

[0072] S604, obtain the frequency domain representation of the light intensity distribution image based on the frequency domain representation of the target grayscale image and the target Gaussian kernel function.

[0073] S606, Perform an inverse Fourier transform on the frequency domain representation of the light intensity distribution image to obtain the light intensity distribution image.

[0074] In some embodiments, the initial image can be a three-channel RGB image, which can be converted into a single-channel grayscale image using a weighted averaging method to obtain the target grayscale image. After obtaining the target grayscale image, the frequency domain representation of the light intensity distribution image can be calculated using the following formula:

[0075] Where I is the frequency domain representation of the initial image, G is the target Gaussian kernel function, and H(u,v) is the frequency domain representation of the light intensity distribution image. Performing an inverse Fourier transform yields:

[0076] in, Represents the inverse Fourier transform. This represents the pixel value of the light intensity at coordinate (x, y) in the light intensity distribution image z. As mentioned earlier, the Gaussian kernel size parameter σ can affect the imaging effect of the light intensity distribution image. As shown in Figures 7(a) to 7(d), which are the initial image, the light intensity distribution image obtained with σ=13, the light intensity distribution image obtained with σ=40, and the light intensity distribution image obtained with σ=100, respectively, it can be seen that as the Gaussian kernel size parameter σ increases, the image gradually becomes more blurred, the details and textures of the target silicon wafer itself gradually decrease, and the uneven distribution trend of light intensity in the initial image becomes more obvious.

[0077] In this embodiment, the initial image is transformed into its frequency domain representation using Fourier transform, which can transform spatial domain convolution calculation into frequency domain dot product calculation, thereby improving the computational performance when processing large-size Gaussian kernels, solving the computational bottleneck in large-scale image processing, and improving the computational efficiency and applicability of image processing.

[0078] As mentioned earlier, various preprocessing operations can be performed on the initial image before low-pass filtering. The following describes an exemplary preprocessing operation.

[0079] In optional embodiments, such as Figure 8 As shown, step S600 includes: S800, obtain the initial grayscale image corresponding to the initial image.

[0080] S802, the boundary of the initial grayscale image is extended by pixels to obtain the target grayscale image.

[0081] In some embodiments, pixel extension can be performed through mirror symmetry extension (extending boundary pixels in a mirror manner), repetitive extension (extending boundary pixel values ​​outward by copying them), or constant padding (padding with fixed gray values ​​such as 0 or 128). The extension width can be a fixed pixel value or determined according to parameters such as the Gaussian kernel size. In some embodiments, the pixel size of the initial grayscale image can also be extended to an integer power of 2 to facilitate subsequent Fourier transforms. Figures 9(a) to 9(c) show a dark grayscale image, a bright grayscale image, and another dark grayscale image after extension, respectively.

[0082] In this embodiment, by extending the boundaries of the initial grayscale image by pixels, a sufficient buffer area can be provided for the subsequent Fourier transform, avoiding boundary wrapping caused by periodic convolution. This improves the accuracy and continuity of light intensity estimation in the effective area of ​​the final obtained light intensity distribution image, thereby enhancing the processing quality of image processing.

[0083] Step S204 The initial image is processed based on the light intensity distribution image to remove the stitching marks of the multiple field images, thereby obtaining the target image corresponding to the initial image.

[0084] In some embodiments, the pixel values ​​at corresponding positions in the initial image and the light intensity distribution image can be directly divided to obtain the pixel values ​​at corresponding positions in the target image. Alternatively, the splicing marks can be removed by using the light intensity distribution image as a gain matrix and performing dynamic brightness normalization on each pixel of the initial image. During processing, different compensation parameters can be set based on features such as local contrast or global grayscale distribution of the initial image. In other embodiments, the corrected target image can be locally smoothed or noise suppressed based on the distribution gradient of the non-uniform light intensity. Figure 10(a) is an exemplary initial image acquired from a target silicon wafer, Figure 10(b) is the target image corresponding to the initial image in Figure 10(a); Figure 10(c) is an exemplary initial image acquired from another target silicon wafer, and Figure 10(d) is the target image corresponding to the initial image in Figure 10(c). It can be seen that a large number of splicing marks in Figure 10(a) have been corrected in Figure 10(b), and a large number of splicing marks in Figure 10(c) have also been corrected in Figure 10(d).

[0085] In this embodiment, the initial image is processed using a light intensity distribution image to compensate for grayscale differences caused by uneven illumination. This eliminates stitching artifacts caused by variations in light intensity at the stitching points of multiple field-of-view images, resulting in a target image with continuous grayscale distribution and no stitching artifacts. This improves the image quality of the target image and the processing efficiency of the image processing.

[0086] As mentioned above, the initial image can be processed in various ways. The following provides an exemplary processing method.

[0087] In an optional embodiment, step S204 includes: Divide the initial pixel value at each pixel position in the initial image by the light intensity pixel value at the corresponding pixel position in the light intensity distribution image to obtain the normalized pixel value. The target pixel value at the corresponding pixel position in the target image is obtained based on the normalized pixel value and the preset scaling factor.

[0088] In practical applications, for coordinates located in the initial image The pixel at that location, its initial pixel value Target pixel value and light intensity distribution coefficient The following relationship should be satisfied:

[0089] Among them, the light intensity distribution coefficient This represents the spatial variation ratio of light intensity on the surface of the target silicon wafer.

[0090] In some embodiments, applying a low-pass filter to the initial image can be equivalent to introducing a hypothetical uniform surface (such as a standard white or silver correction plate) into the imaging model to estimate the non-uniform distribution of the current light field. The light intensity distribution image obtained through this low-pass filtering is then... This represents the actual imaging result of the hypothetical uniform surface under the current light field. Since there is a difference in reflectivity between the hypothetical uniform surface and the target silicon wafer surface, a preset scaling factor k can be introduced to establish... and Relationship:

[0091] The preset scaling factor k can be set according to the actual material properties of the target silicon wafer surface, and is used to correct the difference in reflectivity between the assumed uniform surface and the target silicon wafer surface.

[0092] Meanwhile, for coordinates located in the initial image The pixel at that location, its initial pixel value Light intensity pixel value and normalized pixel values The following relationship should be satisfied:

[0093] Among them, normalized pixel values This represents the coordinates of the target silicon wafer surface. The inherent reflective properties of the splice are no longer affected by sudden changes in lighting at the splice point.

[0094] Further deduction reveals that:

[0095] That is, the target pixel value It can be derived from normalized pixel values and preset scaling factor Multiplying them together, the three satisfy the following relationship:

[0096] Before performing the division operation, the pixel values ​​of the initial image and the light intensity distribution image can be converted from integer to floating-point. In some embodiments, a minimum threshold can be set for the light intensity pixel value. When the light intensity pixel value of a pixel in the light intensity distribution image is lower than the threshold, its corresponding initial pixel value or the set minimum threshold can be directly used as the target pixel value.

[0097] In this embodiment, the normalized result is obtained by first dividing each pixel value of the initial image by the pixel value of the corresponding light intensity distribution image, and then multiplying by a preset scaling factor to recover the target image, thus achieving adaptive correction of the stitching marks from the rotational scanning. In the light intensity distribution estimation stage, the frequency domain Gaussian low-pass filter utilizes the rotational symmetry property of the Gaussian kernel to apply uniform attenuation to the frequency components in any direction in the image, thereby extracting the global low-frequency non-uniform illumination introduced by the imaging system without selective filtering based on the geometric orientation of the stitching marks. In the image correction stage, the pixel-by-pixel division operation only depends on the numerical ratio between the initial pixel value and the corresponding light intensity pixel value at that point, and does not require knowledge of whether the pixel is located on the stitching seam or the specific direction of the stitching seam. This embodiment transforms stitching mark correction into a global illumination compensation problem. It obtains direction-independent light intensity distribution estimates through isotropic filtering, and then uses direction-independent pixel-by-pixel normalization to eliminate the influence of illumination. This means that regardless of the direction of stitching marks in the initial image (including radial, tangential, or random angles generated by circular scanning), there is no need to adjust the algorithm parameters or change the processing path. Consistent mark elimination in all directions can be achieved simply through unified computation. This improves the adaptability and processing flexibility for stitching marks in any direction while ensuring the simplicity of the algorithm.

[0098] After acquiring the target image, various post-processing operations can be performed to improve its quality. An exemplary post-processing operation is described below.

[0099] In an optional embodiment, the method further includes: The target image is subjected to grayscale linear processing to adjust the grayscale range of the target image to a preset grayscale range.

[0100] In some embodiments, the preset grayscale range can be [0, 255]. When the sampling bit depth of the initial image is higher than 8 bits (such as 10-bit, 12-bit, or 16-bit), the preset grayscale range can also be adjusted to [0, 1023], [0, 4095], or [0, 65535]. If the initial image has undergone the aforementioned expansion and extension operations, the grayscale range can also be adjusted for the image region after removing the extended region.

[0101] In this embodiment, the grayscale range of the target image is mapped to a preset grayscale range by linear stretching. While maintaining the image texture details and splicing trace correction effect, the image brightness and contrast can be enhanced, making the defect features on the surface of the target silicon wafer more clearly identifiable. This balances the functionality of splicing trace elimination with the visibility requirements of image quality.

[0102] To make this application easier to understand, the following is combined with... Figure 11 An example application is provided. Wherein: S1. Calculate the circular scanning path based on the silicon wafer diameter and the camera's field of view width, drive the camera to acquire multiple field-of-view images at a preset overlap rate, and stitch them together to form an initial image. Acquire the initial grayscale image corresponding to the initial image, and extend its surrounding boundaries by pixels.

[0103] S2. Based on the fineness of the silicon wafer surface texture, select a Gaussian kernel size parameter σ of 15 to generate a target Gaussian kernel function that can filter high-frequency details and retain low-frequency background information.

[0104] S3. Use Fourier transform to convert the target grayscale image to the frequency domain, and multiply it with the generated frequency domain Gaussian kernel to extract background information by filtering out high-frequency components.

[0105] S4. Perform an inverse Fourier transform on the filtered frequency domain representation to restore it to a spatial domain image, thereby obtaining an image that can characterize the light intensity distribution caused by uneven illumination or imaging system.

[0106] S5. Divide the initial gray value of each pixel position in the initial image by the pixel value of the corresponding position in the light intensity distribution image to eliminate the splicing marks caused by rotating the mosaic.

[0107] S6. Calculate the grayscale range of the corrected image, stretch it to the preset range of [0, 255] through linear mapping, and output a silicon wafer image with clear contrast and no stitching marks.

[0108] Example 2 Figure 12 The diagram schematically illustrates a block diagram of an image processing apparatus according to Embodiment 2 of this application. This apparatus can be divided into one or more program modules. One or more program modules are stored in a storage medium and executed by one or more processors to complete the embodiments of this application. The program modules referred to in the embodiments of this application are a series of computer program instruction segments capable of performing specific functions. The following description will specifically introduce the functions of each program module in this embodiment. For example... Figure 12 As shown, the device 1000 may include: an acquisition module 1100, a processing module 1200, and a removal module 1300, wherein: The acquisition module 1100 is used to acquire an initial image of the target silicon wafer, wherein the initial image is composed of multiple field images stitched together, and one field image is obtained through a single scan; Processing module 1200 is used to perform low-pass filtering on the initial image to obtain a light intensity distribution image corresponding to the initial image, wherein the light intensity distribution image is used to characterize the non-uniform light intensity distribution on the initial image; The removal module 1300 is used to process the initial image according to the light intensity distribution image to remove the stitching marks of the multiple field images and obtain the target image corresponding to the initial image.

[0109] As an optional embodiment, the acquisition module 1100 is further configured to: The number of circular scans and the path length of each circular scan are determined based on the diameter of the target silicon wafer and the width of the camera's field of view. Multiple scan coordinates are determined based on the path length of each circular scan, the length of the camera's field of view, and the preset overlap range; The target silicon wafer is scanned by a camera according to the multiple scanning coordinates to obtain the multiple field-of-view images; The multiple field-of-view images are stitched together according to the preset overlap range to obtain the initial image.

[0110] As an optional embodiment, the processing module 1200 is further configured to: Based on the imaging quality of the initial image, determine the Gaussian kernel size parameter; Based on the Gaussian kernel size parameters, obtain the target Gaussian kernel function; The initial image is low-pass filtered using the target Gaussian kernel function; The target Gaussian kernel function is used to filter high-frequency texture information in the initial image and retain low-frequency light intensity information in the initial image.

[0111] As an optional embodiment, the processing module 1200 is further configured to: Obtain the target grayscale image corresponding to the initial image; Perform a Fourier transform on the target grayscale image to obtain its frequency domain representation; The frequency domain representation of the light intensity distribution image is obtained based on the frequency domain representation of the target grayscale image and the target Gaussian kernel function; The light intensity distribution image is obtained by performing an inverse Fourier transform on its frequency domain representation.

[0112] As an optional embodiment, the processing module 1200 is further configured to: Obtain the initial grayscale image corresponding to the initial image; The boundaries of the initial grayscale image are extended by pixels to obtain the target grayscale image.

[0113] As an optional embodiment, the processing module 1200 is further configured to: Divide the initial pixel value at each pixel position in the initial image by the light intensity pixel value at the corresponding pixel position in the light intensity distribution image to obtain the normalized pixel value. The target pixel value at the corresponding pixel position in the target image is obtained based on the normalized pixel value and the preset scaling factor.

[0114] As an optional embodiment, the device 1000 further includes a grayscale adjustment module for: The target image is subjected to grayscale linear processing to adjust the grayscale range of the target image to a preset grayscale range.

[0115] Example 3 Figure 13 This illustration schematically depicts the hardware architecture of a computer device 10000 suitable for implementing an image processing method according to Embodiment 3 of this application. In some embodiments, the computer device 10000 may be a terminal device such as a smartphone, wearable device, tablet computer, personal computer, in-vehicle terminal, game console, virtual machine, workbench, digital assistant, set-top box, or robot. In other embodiments, the computer device 10000 may be a rack server, blade server, tower server, or cabinet server (including standalone servers or server clusters composed of multiple servers), etc. Figure 13 As shown, the computer device 10000 includes, but is not limited to: a memory 10010, a processor 10020, and a network interface 10030 that can communicate and be linked with each other via a system bus. Wherein: The memory 10010 includes at least one type of computer-readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10010 may be an internal storage module of a computer device 10000, such as the hard disk or memory of the computer device 10000. In other embodiments, the memory 10010 may also be an external storage device of the computer device 10000, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 10000. Of course, the memory 10010 may also include both the internal storage module and the external storage device of the computer device 10000. In this embodiment, the memory 10010 is typically used to store the operating system and various application software installed on the computer device 10000, such as program code for image processing methods. Furthermore, the memory 10010 can also be used to temporarily store various types of data that have been output or will be output.

[0116] In some embodiments, processor 10020 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other chip. Processor 10020 is typically used to control the overall operation of computer device 10000, such as performing control and processing related to data interaction or communication with computer device 10000. In this embodiment, processor 10020 is used to run program code stored in memory 10010 or process data.

[0117] Network interface 10030 may include a wireless network interface or a wired network interface, which is typically used to establish a communication link between computer device 10000 and other computer devices. For example, network interface 10030 is used to connect computer device 10000 to an external terminal via a network, establishing a data transmission channel and communication link between computer device 10000 and the external terminal. The network may be an intranet, the Internet, Global System for Mobile Communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, or other wireless or wired networks.

[0118] It should be pointed out that, Figure 13 Only computer devices with components 10010-10030 are shown; however, it should be understood that it is not required to implement all of the shown components, and more or fewer components may be implemented instead.

[0119] In this embodiment, the image processing method stored in memory 10010 can be further divided into one or more program modules and executed by one or more processors (such as processor 10020) to complete the embodiments of this application.

[0120] Example 4 This application also provides a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the image processing method in the embodiments.

[0121] In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the computer-readable storage medium may be an internal storage unit of a computer device, such as the hard disk or memory of the computer device. In other embodiments, the computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided on the computer device. Of course, the computer-readable storage medium may include both the internal storage unit and the external storage device of the computer device. In this embodiment, the computer-readable storage medium is typically used to store the operating system and various application software installed on the computer device, such as the program code of the image processing method in the embodiment. In addition, the computer-readable storage medium can also be used to temporarily store various types of data that have been output or will be output.

[0122] Example 5 This application also provides a computer program product, including a computer program that, when executed by a processor, implements the methods described in the above embodiments.

[0123] Obviously, those skilled in the art should understand that the modules or steps of the embodiments of this application described above can be implemented using general-purpose computer devices. They can be centralized on a single computer device or distributed across a network of multiple computer devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computer device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of this application are not limited to any particular combination of hardware and software.

[0124] It should be noted that the above are merely preferred embodiments of this application and do not limit the scope of patent protection of this application. Any equivalent structural or procedural changes made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of this application.

Claims

1. An image processing method, characterized in that, The method includes: An initial image of the target silicon wafer is acquired, the initial image being composed of multiple field-of-view images stitched together, and one of the field-of-view images is obtained through a single scan; The initial image is subjected to low-pass filtering to obtain a light intensity distribution image corresponding to the initial image, and the light intensity distribution image is used to characterize the non-uniform light intensity distribution on the initial image. The initial image is processed based on the light intensity distribution image to remove stitching marks from the multiple field-of-view images, thereby obtaining the target image corresponding to the initial image.

2. The method according to claim 1, characterized in that, Obtain an initial image of the target silicon wafer, including: The number of circular scans and the path length of each circular scan are determined based on the diameter of the target silicon wafer and the width of the camera's field of view. Multiple scan coordinates are determined based on the path length of each circular scan, the length of the camera's field of view, and the preset overlap range; The target silicon wafer is scanned by a camera according to the multiple scanning coordinates to obtain the multiple field-of-view images; The multiple field-of-view images are stitched together according to the preset overlap range to obtain the initial image.

3. The method according to claim 1, characterized in that, The initial image is subjected to low-pass filtering, including: Based on the imaging quality of the initial image, determine the Gaussian kernel size parameter; Based on the Gaussian kernel size parameters, obtain the target Gaussian kernel function; The initial image is low-pass filtered using the target Gaussian kernel function; The target Gaussian kernel function is used to filter high-frequency texture information in the initial image and retain low-frequency light intensity information in the initial image.

4. The method according to claim 3, characterized in that, The initial image is low-pass filtered using the target Gaussian kernel function, including: Obtain the target grayscale image corresponding to the initial image; Perform a Fourier transform on the target grayscale image to obtain its frequency domain representation; The frequency domain representation of the light intensity distribution image is obtained based on the frequency domain representation of the target grayscale image and the target Gaussian kernel function; The light intensity distribution image is obtained by performing an inverse Fourier transform on its frequency domain representation.

5. The method according to claim 4, characterized in that, Obtaining the target grayscale image corresponding to the initial image includes: Obtain the initial grayscale image corresponding to the initial image; The boundaries of the initial grayscale image are extended by pixels to obtain the target grayscale image.

6. The method according to claim 1, characterized in that, Processing the initial image based on the light intensity distribution image includes: Divide the initial pixel value at each pixel position in the initial image by the light intensity pixel value at the corresponding pixel position in the light intensity distribution image to obtain the normalized pixel value. Based on the normalized pixel value and the preset scaling factor, the target pixel value at the corresponding pixel position in the target image is obtained.

7. The method according to claim 1, characterized in that, The method further includes: The target image is subjected to grayscale linear processing to adjust the grayscale range of the target image to a preset grayscale range.

8. An image processing apparatus, characterized in that, The device includes: An acquisition module is used to acquire an initial image of the target silicon wafer, wherein the initial image is composed of multiple field-of-view images stitched together, and one field-of-view image is obtained through a single scan; The processing module is used to perform low-pass filtering on the initial image to obtain a light intensity distribution image corresponding to the initial image, wherein the light intensity distribution image is used to characterize the non-uniform light intensity distribution on the initial image; The removal module is used to process the initial image based on the light intensity distribution image to remove the stitching marks of the multiple field images and obtain the target image corresponding to the initial image.

9. A computer device, characterized in that, include: At least one processor; and A memory communicatively connected to the at least one processor; wherein: The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 7.

Images