Image processing method, electronic device, storage medium and computer program product

Abstract

The present application discloses an image processing method, an electronic device, a storage medium and a computer program product. The method comprises comparing the similarity between an input image and a parent node image at a parent node. The method further comprises transferring the input image to a corresponding first-level child node among a plurality of first-level child nodes of the parent node according to the similarity comparison result, wherein the plurality of first-level child nodes respectively correspond to a plurality of similarity ranges. The method further comprises determining a destination node of the input image according to the characteristics of the corresponding first-level child node.

Description

Technical Field

[0001] The present disclosure relates generally to integrated circuits, and more particularly to methods, models, electronic devices, storage media, and computer program products for image processing. Background Art

[0002] Integrated circuit manufacturing is a very precise process involving many process flows. During these process flows, a large number of images related to the integrated circuit manufacturing process are generated, for example, for testing or measuring during the manufacturing process of the integrated circuit for quality control.

[0003] Take the photolithography process as an example. The photolithography process is a key process in the integrated circuit manufacturing process. Through the photolithography process, the graphics required by the integrated circuit design can be realized on the wafer. In the photolithography process, the design graphics need to be prepared on the mask first, and then the mask image is prepared on the wafer. During the photolithography process, there are usually mask images and measurement images obtained by taking pictures of the wafer after photolithography. By collecting a large number of mask images and measurement images, the quality control of the photolithography process can be carried out. Summary of the invention

[0004] According to an exemplary embodiment of the present disclosure, an image processing scheme is provided.

[0005] In a first aspect of the present disclosure, an image processing method is provided. The method includes comparing the similarity of an input image with a parent node image at a parent node. The method also includes transferring the input image to a corresponding first-level child node among a plurality of first-level child nodes of the parent node according to the similarity comparison result, wherein the plurality of first-level child nodes correspond to a plurality of similarity ranges respectively. The method further includes determining a destination node of the input image according to the characteristics of the corresponding first-level child node.

[0006] In a second aspect of the present disclosure, a model for image processing is provided. The model includes at least one parent node, and a first parent node in the at least one parent node has a corresponding parent node image. The model also includes a plurality of first-level child nodes of the first parent node, and the plurality of first-level child nodes correspond to a plurality of similarity ranges, respectively. The model is configured to: compare the similarity of an input image with a parent node image at a first parent node; based on the similarity comparison result, transfer the input image to a corresponding first-level child node in a plurality of first-level child nodes of the first parent node; and determine a destination node of the input image based on the features of the corresponding first-level child node.

[0007] In a third aspect of the present disclosure, an electronic device is provided, comprising: one or more processors; and a storage device for storing one or more programs, which, when the one or more programs are executed by the one or more processors, enables the electronic device to execute the method according to the first aspect of the present disclosure.

[0008] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method according to the first aspect of the present disclosure is implemented.

[0009] In a fifth aspect of the present disclosure, a computer program product is provided, which includes a program code, and when the program code is executed by a processor, the method according to the first aspect of the present disclosure is implemented.

[0010] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0012] Figure 1 A schematic diagram illustrating an example architecture in which various embodiments of the present disclosure can be implemented;

[0013] Figure 2 A flowchart illustrating an example method for image processing according to some embodiments of the present disclosure is shown;

[0014] Figure 3 A schematic diagram showing a model for image processing according to an embodiment of the present disclosure;

[0015] Figure 4A An exemplary schematic diagram showing a time domain image;

[0016] Figure 4B Shown with Figure 4A Schematic diagram of the frequency domain images corresponding to the time domain images in ;

[0017] Figure 5A A schematic diagram showing an arrangement of a mask image based on a tile according to an embodiment of the present disclosure is shown;

[0018] Figure 5B Show Figure 5A A schematic classification result after applying the image processing method according to an embodiment of the present disclosure to a set of images in an arrangement of;

[0019] Fig. 6A A schematic image showing a defect point;

[0020] Figure 6B shows a schematic classification result after applying the image processing method according to an embodiment of the present disclosure to a set of images to be detected;

[0021] Figure 7 A flowchart illustrating an example image processing method according to some embodiments of the present disclosure;

[0022] Figure 8 shows a schematic classification result after applying the image processing method according to an embodiment of the present disclosure to a set of sample images according to an embodiment of the present disclosure;

[0023] Fig. 9 A block diagram of an electronic device capable of implementing an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0024] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0025] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0026] As briefly mentioned above, a large number of images can be generated during the manufacturing process of integrated circuits. By effectively processing and utilizing these large numbers of images, process parameters can be further optimized, thereby improving process stability and reliability, and improving production efficiency and finished product quality.

[0027] The number of images generated during the integrated circuit manufacturing process is huge, and since most of the layers in the process are repeated, there are also a large number of repeated or similar images in the collected images. Therefore, the collected images need to be processed for subsequent use.

[0028] The processing performed on the collected images may include classification processing based on similarity, that is, the images are classified according to different similarity ranges with the reference image, and images with similarity close to the reference image (for example, similarity above 90%) are used as similar images of the reference image. Through the above classification processing, one or more images in a category of images can be conveniently selected for subsequent processing during subsequent use, so that images to be processed can be obtained more specifically, the number of images to be processed can be reduced, and the efficiency of image processing can be improved.

[0029] Currently, in the process of classifying images according to similarity, pairwise matching is usually used. That is, similarity matching is performed between every two images. Since the similarity of images cannot be transferred (for example, the similarity between image A and image B is 80%, and the similarity between image B and image C is 80%, but this does not mean that the similarity between image A and image C is 100%), the pairwise matching of images still requires similarity matching for every two images. In the case of a large number of images, this consumes a huge amount of time and energy.

[0030] In the prior art, in order to reduce the number of images to be processed, a certain number of images are usually selected from the collected images by random sampling. However, the random sampling method may lose some important attributes of the image to some extent. In addition, the randomly sampled data set also conforms to the original data distribution. If the original data set is unbalanced, then the sampled data set will usually maintain the imbalance of the original data set, resulting in unsatisfactory results in the subsequent processing of the sampled images.

[0031] Therefore, there is an urgent need for an image processing method (for example, similarity-based image classification processing) that can improve processing time, processing efficiency, and processing quality, so as to provide high-quality processed images for integrated circuit manufacturing, thereby further improving manufacturing efficiency and finished product quality.

[0032] According to an embodiment of the present disclosure, a method, model, electronic device, computer-readable storage medium and computer program product for image processing are proposed. The image processing method according to an embodiment of the present disclosure includes: comparing the similarity of an input image with a parent node image at a parent node; according to the similarity comparison result, transferring the input image to a corresponding first-level child node among a plurality of first-level child nodes of the parent node, wherein the plurality of first-level child nodes correspond to a plurality of similarity ranges respectively; and determining a destination node of the input image according to the characteristics of the corresponding first-level child node.

[0033] According to the method for image processing proposed herein, the speed of classifying images based on similarity can be greatly improved, the image processing time and complexity can be reduced, and the quality and rate of image classification processing can be improved. Therefore, the solution disclosed in the present invention can provide high-quality and more targeted images for subsequent further processing of the images.

[0034] Figure 1 1 shows an example environment 100 in which various embodiments of the present disclosure can be implemented. Figure 1 As shown, the electronic device 110 may receive a set 102 of input images. In some embodiments, the input images in the set 102 of input images may include various types of images collected during the integrated circuit manufacturing process. For example, the input images may include one or more of a mask image, a measurement image obtained by photographing a wafer after photolithography imaging. In some embodiments, the input images may include frequency domain images obtained by frequency domain conversion of various types of time domain images collected during the integrated circuit manufacturing process, for example, the input images may include frequency domain images of mask images, frequency domain images of measurement images, etc.

[0035] In some embodiments, the wafer after photolithography imaging can be photographed by a measuring device such as a scanning electron microscope (SEM) to obtain a measurement image. In addition, the measurement image of the wafer surface can also be obtained by any other appropriate measuring device through various appropriate technical means, and the present disclosure does not limit the specific method of obtaining the measurement image.

[0036] In some embodiments, the electronic device 110 can construct a model (e.g., a screening processor) for processing images (e.g., classifying images based on similarity), and the image processing model can be a tree structure and include multiple levels, and the nodes of each level can serve as the parent nodes of the next level, and accordingly, the nodes of each level can serve as the child nodes of the previous level. According to an embodiment of the present disclosure, the electronic device 110 can receive each input image in the set 102 of input images. The electronic device 110 can compare the similarity of the input image with the parent node image at the parent node. The electronic device 110 can pass the input image to the corresponding first-level child node of the multiple first-level child nodes of the parent node according to the similarity comparison result, wherein the multiple first-level child nodes correspond to multiple similarity ranges respectively. The electronic device 110 can further determine the destination node of the input image, that is, the next node to which the input image is passed, based on the characteristics of the corresponding first-level child node.

[0037] According to the method for image processing of the embodiment of the present disclosure, the speed of the electronic device to classify and process images based on similarity can be greatly improved, the processing time and complexity can be reduced, and the quality and rate of image classification processing can be improved. In addition, by executing the above-mentioned image processing method on the input image, the electronic device 110 can associate and save the input image with the nodes of the corresponding level based on the similarity of the image. Thus, it is beneficial for the electronic device 110 to select appropriate images from the classified images for subsequent processing according to the goals and requirements of the subsequent processing.

[0038] The electronic device 110 may be any electronic device with computing capabilities. As a non-limiting example, the electronic device 110 may be any type of fixed electronic device, mobile electronic device, or portable electronic device, including but not limited to a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a multimedia computer, a mobile phone, a smart home device, a wearable electronic device, etc. In some embodiments, all or part of the components of the electronic device 110 may be distributed in the cloud. The present disclosure does not limit the specific type of the electronic device 110.

[0039] Figure 2 FIG. 2 is a flowchart of an exemplary method 200 for image processing according to some embodiments of the present disclosure. Figure 1 Method 200 is described. Figure 2 The method 200 in can be Figure 1 In addition, the numbers in the flowchart do not represent the order in which these steps are executed. Some or all of these steps can be executed in parallel, or the execution order can be interchanged, and this is not limited in the present disclosure.

[0040] In box 202, the electronic device 110 may compare the input image with the parent node image at the parent node for similarity. The input image may be an input image in the set 102 of input images. In some embodiments, the input image may include various types of images collected during the integrated circuit manufacturing process. For example, the input image may include one or more types of images such as a mask image, a measurement image obtained by photographing a wafer after photolithography imaging, etc. In some embodiments, the input image may include a frequency domain image after frequency domain conversion of various types of time domain images collected during the integrated circuit manufacturing process, for example, the input image may include a frequency domain image of a mask image, a frequency domain image of a measurement image, etc.

[0041] In some embodiments, a model (e.g., a screening processor) for processing images (e.g., classifying images based on similarity) can be constructed in the electronic device 110. The model can be a tree structure and include multiple levels. The nodes of each level can serve as the parent nodes of the next level, and accordingly, the nodes of each level can be the child nodes of the previous level.

[0042] Figure 3 FIG. 3 is a schematic diagram of a model 300 for image processing according to an embodiment of the present disclosure. Figure 3 As shown, Figure 3 The model 300 in includes a root node 310. The model 300 according to an embodiment of the present disclosure may include child nodes of multiple levels. Figure 3 The model 300 in FIG. 3 illustrates three levels of sub-nodes: a first level sub-node 320, a second level sub-node 330, and a third level sub-node 340. The sub-nodes of each level may include multiple sub-nodes of the corresponding level, and each sub-node may correspond to a corresponding similarity range.

[0043] by Figure 3 Taking the model 300 in FIG. 3 as an example, the first-level subnode 320 includes 10 subnodes, and the 10 subnodes correspond to the similarity ranges 0~0.1, 0.1~0.2, 0.2~0.3, 0.3~0.4, 0.4~0.5, 0.5~0.6, 0.6~0.7, 0.7~0.8, 0.8~0.9, and 0.9~1. Similarly, the second-level subnode 330 and the third-level subnode 340 also include 10 subnodes, and the 10 subnodes in each level correspond to the above-mentioned similarity ranges.

[0044] Understandably, Figure 3 The model 300 for image processing in is only schematic. According to needs and actual requirements, the model 300 may include more or fewer levels of sub-nodes, and the sub-nodes in each level may correspond to a similarity range greater than 10 or less than 10. In addition, the similarity range of the sub-nodes in each level may be divided evenly or unevenly. For example, illustratively, the similarity range in each level may be: 0~0.2, 0.2~0.5, 0.5~0.6, 0.6~0.8, 0.8~0.9, 0.9~1, etc. Moreover, the division criteria of the similarity ranges between the levels may be the same or different, and the present disclosure does not limit this. In addition, although the similarity is expressed by decimals here, it is understandable that the method according to the embodiment of the present disclosure may also use other methods (e.g., percentages) to express the similarity, and the present disclosure does not limit this.

[0045] In some embodiments, the electronic device 110 can compare the input image with the parent node image at the parent node for similarity. The electronic device 110 can use various known or future developed technologies to compare the input image with the parent node image at the parent node for similarity. It is understandable that the "parent node image at the parent node" here can refer to an image associated with the parent node. In some embodiments, the image identifier of the parent node image can be stored in association with the parent node. Through the image identifier, the electronic device 110 can call the parent node image from the storage device and compare the called parent node image with the input image for similarity. In some embodiments, the parent node image can also be stored in association with the parent node.

[0046] At box 204, the electronic device 110 may transfer the input image to a corresponding first-level child node among a plurality of first-level child nodes of the parent node according to the similarity comparison result, wherein the plurality of first-level child nodes correspond to a plurality of similarity ranges, respectively.

[0047] In some embodiments, the electronic device 110 may transfer the input image to a first-level child node corresponding to the similarity comparison result among multiple first-level child nodes of the parent node based on the similarity comparison result. For example, when the similarity comparison result is 0.25, the electronic device 110 may transfer the input image to a first child node corresponding to a similarity range of 0.2 to 0.3 among multiple first-level child nodes of the parent node. In other words, the electronic device 110 classifies the similarity comparison result of the input image into the child nodes of the corresponding similarity range. It is understandable that, for the case where the similarity comparison result is equal to the endpoint value of the similarity range, the input image may be classified into one or more child nodes whose similarity range has an endpoint value equal to the similarity comparison result, based on need or demand. For example, when the similarity comparison result is 0.2, the electronic device 110 may pass the input image to the first child node in the first-level child node of the parent node corresponding to the similarity range of 0.2~0.3, or may pass the input image to the first child node in the first-level child node of the parent node corresponding to the similarity range of 0.1~0.2, or may pass the input image to the first child node in the first-level child node of the parent node corresponding to the similarity range of 0.2~0.3 and the first child node corresponding to the similarity range of 0.1~0.2, and the present disclosure is not limited to this.

[0048] In box 206, the electronic device 110 can determine the destination node of the input image based on the characteristics of the corresponding first-level child nodes. In some embodiments, the characteristics of the child nodes can include at least one of the following items: the similarity range corresponding to the child node, or whether the child node has a corresponding child node image (i.e., whether the child node has an associated child node image). The electronic device 110 can determine the destination node of the input image based on the characteristics of the child node.

[0049] In some embodiments, the corresponding first-level child node may have a corresponding first-level child node image (e.g., a child node image associated with the corresponding first-level child node). In response to the corresponding first-level child node having a corresponding first-level child node image, the electronic device 110 may compare the input image with the corresponding first-level child node image for similarity. The electronic device 110 may, based on the similarity comparison result, pass the input image to a corresponding second-level child node among a plurality of second-level child nodes of the corresponding first-level child node. In some embodiments, the plurality of second-level child nodes are child nodes of the corresponding first-level child node, and correspond to a plurality of similarity ranges, respectively. In some embodiments, the implementation of the electronic device 110 passing the input image to a corresponding second-level child node among a plurality of second-level child nodes of the corresponding first-level child node based on the similarity comparison result can be understood with reference to box 204, and for the sake of simplicity, it will not be repeated here.

[0050] In some embodiments, the corresponding first-level child node may not have a corresponding first-level child node image (for example, the corresponding first-level child node does not have an associated child node image). In response to the corresponding first-level child node not having a corresponding first-level child node image, the electronic device 110 may associate the input image with the corresponding first-level child node and save it as the corresponding first-level child node image (that is, use the input image as a child node image associated with the corresponding first-level child node). In some embodiments, the electronic device 110 may associate the image identifier of the input image with the first-level child node and save it, so that the input image is associated with the first-level child node as the first-level child node image at the first-level child node. In other embodiments, the electronic device 110 may associate other identification information used to identify the input image with the first-level child node and save it.

[0051] In some embodiments, the similarity range corresponding to the corresponding first-level child node may have a maximum endpoint value among the endpoint values ​​of multiple similarity ranges. In response to the input image being transferred to the first-level child node having the maximum endpoint value in the similarity range in box 204, the electronic device 110 associates and saves the input image with the corresponding first-level child node (i.e., associates the input image with the corresponding first-level child node), indicating that the input image is a similar image of the parent node.

[0052] It is understandable that after executing the steps in box 206 and passing the input image to the destination node, the electronic device 110 can continue to iteratively execute the steps in boxes 204 and 206 for the input image passed to the destination node until the input image reaches the lowest level child node of the model 300 or is associated with a child node in the child nodes of a certain level (for example, a child node with a similarity range with a maximum endpoint value) and is saved. In some embodiments, the electronic device 110 can continuously receive input images from the set 102 of input images, thereby performing the image processing method according to an embodiment of the present disclosure on the input image.

[0053] According to the image processing method of the embodiment of the present disclosure, the speed at which the electronic device 110 performs classification processing on images based on similarity can be greatly improved, the processing time and complexity can be reduced, and the quality and rate of image processing can be improved. In addition, by performing the above-mentioned image processing method on the input image, the electronic device 110 can associate and save the input image with the sub-nodes of the corresponding level based on the similarity between the input image and the reference image, thereby facilitating the electronic device 110 to select appropriate images from the classified images for subsequent processing according to the goals and requirements of subsequent processing.

[0054] In addition, the model for image processing (e.g., image screening processor) constructed according to the embodiment of the present disclosure can easily adapt to the concurrency of multiple threads and multiple processes, so as to accelerate data more flexibly and improve the computational reasoning speed of the software process. In addition, the model can have a tree structure, and the tree structure can also be easily saved, so that when a new input image enters, the existing model can be easily expanded. For example, as the number of input images increases, the number of levels of the sub-nodes of the model can be expanded to accommodate similarity-based classification processing for more input images.

[0055] The following will be combined Figure 3 An example process of processing an input image according to an embodiment of the present disclosure is described by way of example.

[0056] like Figure 3 As shown, Figure 3The model (e.g., screening processor) 300 in includes a root node 310. The image associated with the root node 310 (i.e., the root node 310 has a root node image) can be randomly selected from the set of input images 102. The electronic device 110 can randomly select an input image from the set of input images 102 and use the selected input image as the root node image R at the root node 310.

[0057] Figure 3 The model 300 in FIG. 1 illustrates three levels of sub-nodes: a first level sub-node 320, a second level sub-node 330, and a third level sub-node 340. Each level of sub-nodes includes multiple sub-nodes, and each sub-node corresponds to a corresponding similarity range. Figure 3 As shown, the first-level subnode 320 includes 10 subnodes, and the 10 subnodes correspond to similarity ranges of 0~0.1, 0.1~0.2, 0.2~0.3, 0.3~0.4, 0.4~0.5, 0.5~0.6, 0.6~0.7, 0.7~0.8, 0.8~0.9, and 0.9~1. Similarly, the second-level subnode 330 and the third-level subnode 340 also include 10 subnodes, and the 10 subnodes of each level correspond to the above-mentioned similarity ranges.

[0058] The electronic device 110 may receive an input image in the set 102 of input images, such as an input image A. The electronic device 110 may compare the similarity between the input image A and the root node image R at the root node 310. The electronic device 110 may use any appropriate method to compare the similarity between the input image A and the root node image R at the root node 310. For example, if the electronic device 110 determines that the similarity between the input image A and the root node image R is 0.67, the input image A is transferred to the child node n1 corresponding to the similarity range of 0.6 to 0.7 in the child node 320 of the first level, and the child node n1 is a child node of the root node 310.

[0059] The electronic device 110 determines that the similarity range of the child node n1 does not have the maximum endpoint value among the endpoint values ​​of the multiple similarity ranges. And the electronic device 110 can also determine whether the child node n1 has a corresponding child node image I n1 , that is, the associated child node image I n1 .

[0060] In response to the child node n1 having a corresponding child node image I n1 , the electronic device 110 can combine the input image A with the child node image I n1The electronic device 110 performs a similarity comparison, and according to the similarity comparison result, the input image A is further transferred to the child node of the child node n1. That is, the electronic device 110 transfers the input image A to the corresponding child node in the second-level child node 330 according to the similarity comparison result. For example, the electronic device 110 determines that the input image A is similar to the child node image n1. I n1 The similarity is 0.45, then the input image A is passed to the child node n2 corresponding to the similarity 0.4~0.5 in the second-level child node 330, and the child node n2 is the child node of the child node n1.

[0061] In response to the child node n2 having a corresponding child node image I n2 , the electronic device 110 can combine the input image A with the child node image I n2 The electronic device 110 performs a similarity comparison, and according to the similarity comparison result, the input image A is further transferred to the child node of the child node n2. That is, the electronic device 110 transfers the input image A to the corresponding child node in the third-level child node 340. For example, the electronic device 110 determines that the input image A is similar to the child node image n2. I n2 The similarity is 0.82, then the input image A is passed to the child node n3 corresponding to the similarity 0.8~0.9 in the child node 340 of the third level, and the child node n3 is the child node of the child node n2.

[0062] The electronic device 110 may determine that the child node 340 of the third level is the child node of the lowest level of the model 300, that is, the child node 340 of the third level no longer has a child node, and the electronic device 110 determines that the operation for the input image A has been performed. The electronic device 110 may continue to receive the next input image B from the set 102 of input images and perform similar operations as described above.

[0063] In some embodiments, the electronic device 110 may receive an input image B in the set 102 of input images. The electronic device 110 may compare the similarity between the input image B and the root node image R at the root node 310. The electronic device 110 may use any appropriate method to compare the similarity between the input image B and the root node image R. For example, if the electronic device 110 determines that the similarity between the input image B and the root node image R is 0.81, the input image B is transferred to the child node n4 corresponding to the similarity range of 0.8 to 0.9 in the child node 320 of the first level, and the child node n4 is a child node of the root node 310.

[0064] The electronic device 110 determines that the similarity range of the child node n4 does not have the maximum endpoint value among the endpoint values ​​of the multiple similarity ranges. And the electronic device 110 can also determine whether the child node n4 has a corresponding child node image I n4 , that is, the associated child node image I n4 .

[0065] In response to the child node n4 not having a corresponding child node image I n4 , the electronic device 110 may associate the input image B with the child node n4 and save it as the child node image of the child node n4. I n4 For example, the electronic device 110 may store the image identifier of the input image B in association with the child node n4, so that the input image B is associated with the child node n4, and the input image B is used as the child node image at the child node n4. Alternatively, the electronic device 110 may store other identification information for identifying the input image B in association with the child node n4, and the input image B is used as the child node image at the child node n4.

[0066] The electronic device 110 may determine that the operation on the input image B has been completed, and the electronic device 110 may continue to receive the next input image C from the set 102 of input images and perform similar operations as described above.

[0067] In some embodiments, the electronic device 110 may receive an input image C in the set 102 of input images. The electronic device 110 may compare the similarity between the input image C and the root node image R at the root node 310. The electronic device 110 may use any appropriate method to compare the similarity between the input image C and the root node image R at the root node 310. For example, if the electronic device 110 determines that the similarity between the input image B and the root node image R is 0.95, the input image C is transferred to the child node n5 corresponding to the similarity range 0.9~1 in the child node 320 of the first level, and the child node n5 is a child node of the root node 310.

[0068] In response to the similarity range corresponding to the child node n5 having the maximum endpoint value 1 among the endpoint values ​​of the multiple similarity ranges, the electronic device 110 can associate the input image C with the child node n5 corresponding to the similarity range 0.9~1 in the first-level child node 320 and save it, and use the input image C as a similar image of the root node image R. In some embodiments, the electronic device 110 can associate the image identifier of the input image C with the child node n5 and save it, so that the input image C is associated with the child node n5 as the first-level child node image at the child node n5. Alternatively, the electronic device 110 can associate other identification information for identifying the input image C with the child node n5 and save it, so as to use the input image C as the child node image at the child node n5.

[0069] The electronic device 110 may continuously process the input images in the set 102 of input images, thereby associating and storing the processed input images with the corresponding destination nodes. Thus, the processing of the input images in the set 102 of input images may be achieved.

[0070] In addition, it is understood that the above Figure 3 The example process described is only illustrative. According to the comparison result of the input image in the set 102 of input images and the root node image, the input image can be associated with any destination node in the model 300 and saved. In addition, it can be understood that the "parent node" and "child node" in the embodiment of the present disclosure are relative. For example, the parent node Node 1 can be the child node of its parent node Node 2.

[0071] In some embodiments, the node image at each node is a frequency domain image. Due to the characteristics of the mask image, most of the mask images are similar. The pattern between two time domain mask images may be the same, and the only difference is some offsets (such as up and down offsets or left and right offsets). This offset of the time domain image is usually considered to be two different images. If it is regarded as a different image because of the offset, the difference between the data cannot actually be reflected. In this regard, according to the method for image processing of an embodiment of the present disclosure, the time domain image is subjected to a frequency domain transformation (for example, Fourier transform) to obtain a frequency domain image. The converted frequency domain image no longer reflects the offset in the time domain image. When the pattern of the time domain image is the same, even if an offset occurs, the two converted frequency domain images are consistent. It can be considered that the two images are similar. Therefore, by performing a frequency domain conversion on the time domain image, most of the similar images can be filtered out, thereby maintaining the diversity of the image.

[0072] Figure 4A An exemplary schematic diagram showing a time domain image, and Figure 4B Shown with Figure 4ASchematic diagram of the frequency domain images corresponding to the time domain images in . Figure 4A It can be seen that image 1 and image 2 have the same pattern, but Figure 4A The image 2 in the image is shifted downward compared to the image 1, which is also considered as a different image in the time domain.

[0073] Figure 4B The frequency domain images F1 and F2 corresponding to the image 1 and the image 2 are shown in FIG. Figure 4B It can be seen that the frequency domain images F1 and F2 corresponding to the image 1 and the image 2 are basically consistent, thus reflecting that the image 1 and the image 2 are similar images.

[0074] In some embodiments, the node image at each node is a frequency domain image. Accordingly, the input image in the set of input images is a frequency domain image. In some embodiments, the electronic device 110 may perform a frequency domain transform on the image in the set of images to be processed to obtain a corresponding frequency domain image, and use the obtained frequency domain image as the input image (e.g., the set 102 of input images). For example, for the root node image at the root node 310, the electronic device 110 may obtain a first image from the set of images to be processed (e.g., randomly obtain the first image). The electronic device 110 may perform a frequency domain transform on the obtained first image, and use the first frequency domain image after the frequency domain transform as the root node image. Similarly, the electronic device 110 may perform a similar frequency domain conversion operation on other images to be processed, so that the input image input into the model 300 is a frequency domain image, thereby achieving the diversity of the input image. For example, the electronic device 110 may obtain a second image from the set of images to be processed. The electronic device 110 may perform a frequency domain transform on the obtained second image, and input the second frequency domain image after the frequency domain transform into the model 300 as the input image for processing.

[0075] In some embodiments, an image processing method according to an embodiment of the present disclosure may be used in the process of posting back. In the posting back scenario, the time domain image corresponding to the image in the set of input images may include a mask image. These time domain mask images are based on tiles. The parent node image at the parent node in the model for image processing according to an embodiment of the present disclosure may be a frequency domain image, and the time domain image corresponding to the parent node image includes a mask image. The electronic device 110 may determine that the input image is a similar image to the parent image based on the similarity comparison result. The electronic device 110 may perform optical proximity correction (OPC) and verification on the time domain image corresponding to the parent node image to obtain a calibrated time domain mask image. The electronic device 110 may post the calibrated time domain mask image back to the time domain image corresponding to the input image.

[0076] Figure 5A FIG. 1 is a schematic diagram showing an arrangement of a mask image based on a tile according to an embodiment of the present disclosure. For example, Figure 5A The image includes 8 mask images P1 to P8. It can be understood that Figure 5A The arrangement diagrams illustrated in FIG. 1 are exemplary only and are used for illustrative purposes.

[0077] The electronic device 110 may use the image processing method according to the embodiment of the present disclosure to Figure 5A Images P1 to P8 in the arranged image set are frequency-domain transformed to obtain a set of input images, for example, the set of input images may include frequency-domain images P'1 to P'8. The electronic device 110 performs similarity-based classification processing on the set of input images. Figure 5B Show Figure 5A The electronic device 110 can use the frequency domain image P'1 corresponding to the mask image P1 as the root node image. Figure 5B It can be seen that by adopting the image processing method according to the embodiment of the present disclosure, it can be determined that images P'3, P'7 and P'6 have the highest similarity with the root node image P'1, and therefore can be used as similar images of the root node image P'1 (correspondingly, the time domain images P3, P7 and P6 are similar to the time domain image P1). Similarly, P'4 and P'5 have the highest similarity with their parent node image P'2, and therefore can be used as similar images of their parent node image P'2 (correspondingly, the time domain images P4 and P5 are similar to the time domain image P2). It can be understood that in order to ensure accuracy, in the reply scenario, the similarity range of the model can be adjusted, for example, using a similarity range of 0.95~1.

[0078] Therefore, according to Figure 5BAccording to the processing result shown in , the electronic device 110 can perform optical proximity effect correction and verification on the time domain image P1 corresponding to the root node image P'1 and the time domain image P2 corresponding to the parent node image P'2 of the frequency domain images P'4 and P'5, so as to obtain calibrated time domain mask images P 1c and P 2c .

[0079] After obtaining the calibrated time domain mask image P 1c and P 2c Afterwards, based on Figure 5B , P'3, P'7 and P'6 are similar images of the root node image P'1 and P'4 and P'5 are similar images of the parent node image P'2. The electronic device 110 may convert the calibrated time domain mask image P 1c Paste it back to the time domain images P3, P7 and P6 corresponding to the images P'3, P'7 and P'6 respectively, and convert the calibrated time domain mask image P 2c Paste back to the time domain images P4 and P5 corresponding to the images P'4 and P'5 respectively.

[0080] Since the temporal mask image P 1c and P 2c Having been corrected and verified for the optical proximity effect, the number of correction rounds for the mask image can be reduced by pasting it back to a similar image that is similar to it, so that a converged mask image can be obtained after a smaller number of correction rounds.

[0081] In some embodiments, the method for image processing according to the embodiment of the present disclosure may be used in the process of defect repair. In the defect repair scenario, the time domain image corresponding to the image in the set of input images may include an image with defect points. In some embodiments, the parent node image at the parent node in the model for image processing according to the embodiment of the present disclosure is a frequency domain image, and the time domain image corresponding to the parent node image includes an image with defect points.

[0082] The electronic device 110 may determine that the input image is a similar image to the parent node image based on the similarity comparison result. The electronic device 110 may apply the same defect repair rule to the time domain image corresponding to the parent node image and the time domain image corresponding to the input image.

[0083] Fig. 6A A schematic image P showing a defective point D1 and P D2 It is understandable that in the defect repair scenario, there are a large number of images to be detected. For the sake of simplicity, Fig. 6A Two images with defective points are only schematically shown in FIG.

[0084] The electronic device 110 may obtain a set of images to be detected, and perform frequency domain conversion on the images in the set of images to be detected to obtain a set of input images, wherein the set of input images includes frequency domain images. For example, the set of images to be detected received by the electronic device 110 may include time domain images. P D1 ~ P D9 The electronic device 110 may perform frequency domain conversion on the images in the set of images to be detected to obtain a set of input images. For example, the set of input images includes frequency domain images. P D1f ~ P D9f .

[0085] The electronic device 110 may employ the image processing method according to an embodiment of the present disclosure to perform image processing on a set of input images. Figure 6B FIG. 2 shows a schematic classification result after applying the image processing method according to an embodiment of the present disclosure to an image in the set of images to be detected. D1 The corresponding frequency domain image P D1f As the root node image. Figure 6B It can be seen from the figure that by adopting the image processing method according to the embodiment of the present disclosure, the image P can be determined. D4f , P D5f With the root node image P D1f The similarity is the highest, so it can be used as the root node image P D1f . Similarly, P D8f and P D9f With its parent node image P D2f has the highest similarity, so it can be used as the parent node image P D2f Similar images of .

[0086] Therefore, according to Figure 6B As a result, the electronic device 110 can D1f The corresponding time domain image P D1 and node image P D4f , P D5f The corresponding time domain images P D4 , P D5 The same defect repair rules are applied. The electronic device 110 may D2f The corresponding time domain image P D2 and node image P D8f , P D9f The corresponding time domain images P D8 and P D9 The same bug fixing rules apply.

[0087] By applying the image processing method according to the embodiment of the present disclosure in the defect detection scenario, similar images similar to the parent node image can be quickly detected, and the same or similar repair values ​​or repair rules can be applied to the time domain image corresponding to the parent node image and the time domain image corresponding to its similar image, thereby improving the efficiency of the repair and achieving rapid repair.

[0088] In some embodiments, the parent node image at the parent node in the model for image processing according to an embodiment of the present disclosure may be a frequency domain image, and the time domain image corresponding to the parent node image includes a sample image for training a neural network model. For example, a neural network model may be trained to perform OPC calibration in a lithography process, and the corresponding training sample image may be a sample mask image. It is understood that, depending on the purpose of use of the neural network model, a corresponding sample image may be used.

[0089] Figure 7 A flow chart of an example method 700 for image processing is shown according to some embodiments of the present disclosure. Figure 7 The method 700 may be Figure 1 In addition, the numbers in the flowchart do not represent the order in which these steps are executed. Some or all of these steps can be executed in parallel, or the execution order can be interchanged, and this is not limited in the present disclosure.

[0090] In box 702, the electronic device 110 can determine a subset of multiple first-level child nodes, and the endpoint values ​​of the similarity range corresponding to the child node elements in the subset are all less than a predetermined similarity threshold. In some embodiments, the predetermined similarity threshold can be 1. In some embodiments, the predetermined similarity threshold can be 0.9 or 0.8, depending on the requirements for sample data and the number of samples during the training process. For example, when the predetermined similarity threshold is 1, it can be indicated that the electronic device 110 can select the child node image at the child node where the image other than the image with the greatest similarity to the parent node image is located.

[0091] In box 704, the electronic device 110 may obtain the child node image at the child node element in the subset. In box 706, the electronic device 110 may use the time domain image corresponding to the parent node image and the obtained time domain image corresponding to the child node image as sample images for training the neural network model.

[0092] Will combine Figure 8 Describe Figure 7 Implementation process of method 700. Figure 8Schematic classification result 800 after applying the image processing method according to the embodiment of the present disclosure to a set of sample images according to the embodiment of the present disclosure. For the sub-nodes of each level, the electronic device 110 can determine a subset of the sub-nodes in each level, and the endpoint values ​​of the similarity range corresponding to the sub-node elements in the subset are all less than the predetermined similarity threshold. Figure 8 For example, the predetermined similarity threshold may be 1. For the first-level child nodes 811-1 to 811-10, the electronic device 110 may determine that the selected child node subset includes child nodes 811-1 to 811-9. The endpoint values ​​of the similarity ranges corresponding to nodes 811-1 to 811-9 are all less than 1. Similarly, for the second-level child nodes 821-1 to 821-10, ... 829-1 to 829-10, the electronic device 110 may select child nodes where images other than those having a similarity range of 0.9 to 1.0 with the parent node images at the parent nodes 811-1 to 811-9 are located, i.e., determine that the selected child node subset includes child nodes 821-1 to 821-9, ... 829-1 to 829-9, and the endpoint values ​​of the similarity ranges corresponding to the selected child nodes are all less than 1.

[0093] The electronic device 110 may obtain a sub-node image at a sub-node element in the selected sub-node subset (ie, a sub-node image associated with the sub-node element). Figure 8 The time domain image corresponding to the root node image and the time domain image corresponding to the acquired child node image are used as sample images for training the neural network model.

[0094] By adopting the method for image processing according to the embodiment of the present disclosure, in the process of selecting sample images for training a neural network, a large number of similar images can be excluded, and images with diversity can be accurately screened out, thereby improving the training effect of machine learning.

[0095] According to an embodiment of the present disclosure, a model for image processing is also provided. Figure 3 The model 300 shown in the figure can be used to understand the model for image processing according to an embodiment of the present disclosure. In some embodiments, the model may include at least one parent node, and the first parent node in the at least one parent node has a corresponding parent node image. The model may also include multiple first-level child nodes of the first parent node, and the multiple first-level child nodes correspond to multiple similarity ranges respectively. The model is configured to: compare the input image with the parent node image at the first parent node for similarity, and based on the similarity comparison result, pass the input image to the corresponding first-level child node in the multiple first-level child nodes of the first parent node, and determine the destination node of the input image based on the characteristics of the corresponding first-level child node.

[0096] In some embodiments, determining the destination node of an input image based on features of a corresponding first-level sub-node may include: in response to the corresponding first-level sub-node having a corresponding first-level sub-node image, performing a similarity comparison between the input image and the corresponding first-level sub-node image; and based on the similarity comparison result, passing the input image to a corresponding second-level sub-node among a plurality of second-level sub-nodes, wherein the plurality of second-level sub-nodes are sub-nodes of the corresponding first-level sub-node and correspond to a plurality of similarity ranges, respectively.

[0097] In some embodiments, determining the destination node of the input image based on the characteristics of the corresponding first-level child node may include: in response to the corresponding first-level child node not having a corresponding first-level child node image, associating the input image with the corresponding first-level child node and saving it as the corresponding first-level child node image.

[0098] In some embodiments, determining the destination node of an input image based on features of a corresponding first-level child node may include: in response to a similarity range corresponding to the corresponding first-level child node having a maximum endpoint value, associating the input image with the corresponding first-level child node and saving the input image, wherein the input image represents a similar image to the parent node image.

[0099] In some embodiments, the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes a mask image. The model is also configured to: determine that the input image is a similar image of the parent node image based on the similarity comparison result; perform optical proximity effect correction and verification on the time domain image corresponding to the parent node image to obtain a calibrated time domain mask image; and paste the calibrated time domain mask image back to the time domain image corresponding to the input image.

[0100] In some embodiments, the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes an image with defect points. The model is also configured to: determine that the input image is a similar image of the parent node image based on the similarity comparison result; and apply the same defect repair rules to the time domain image corresponding to the parent node image and the time domain image corresponding to the input image.

[0101] In some embodiments, the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes a sample mask image for training a neural network model.

[0102] In some embodiments, the model is also configured to: determine a subset of multiple first-level child nodes, wherein the endpoint values ​​of the similarity ranges corresponding to the child node elements in the subset are all less than a predetermined similarity threshold; obtain child node images at the child node elements in the subset; and use the time domain image corresponding to the parent node image and the time domain image corresponding to the obtained child node image as sample images for training the neural network model.

[0103] In some embodiments, the first parent node includes a root node, and the model is further configured to: obtain a first image from a set of images to be processed; perform a frequency domain transform on the first image; and use the first frequency domain image after the frequency domain transform as a root node image at the root node.

[0104] In some embodiments, the model is further configured to: obtain a second image from the set of images to be processed; perform a frequency domain transform on the second image; and use the second frequency domain image after the frequency domain transform as an input image.

[0105] Fig. 9 1 shows a schematic block diagram of an example device 900 that can be used to implement an embodiment of the present disclosure. The device 900 can be used to implement Figure 1 Electronic device 110. As shown in the figure, device 900 includes a processing unit 901, such as a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) 902 or computer program instructions loaded from a storage unit 908 to a random access memory (RAM) 903. In RAM 903, various programs and data required for the operation of device 900 can also be stored. Processing unit 901, ROM 902 and RAM 903 are connected to each other via a bus 904. Input / output (I / O) interface 905 is also connected to bus 904.

[0106] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0107] The processing unit 901 performs the various methods and processes described above, such as methods 200 and 700. For example, in some embodiments, methods 200 and 700 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the processing unit 901, one or more steps of the methods 200 and 700 described above may be performed. Alternatively, in other embodiments, the processing unit 901 may be configured to perform methods 200 and 700 in any other suitable manner (e.g., by means of firmware).

[0108] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip systems (SOCs), load programmable logic devices (CPLDs), and the like.

[0109] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0110] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0111] In addition, although each operation is described in a specific order, this should be understood as requiring such operation to be performed in the specific order shown or in a sequential order, or requiring that all illustrated operations should be performed to obtain desired results. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination mode.

[0112] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A method for image processing, comprising: Compare the similarity between the input image and the parent node image at the parent node; According to the similarity comparison result, the input image is transferred to a corresponding first-level child node among a plurality of first-level child nodes of the parent node, wherein the plurality of first-level child nodes respectively correspond to a plurality of similarity ranges; as well as Determine the destination node of the input image according to the features of the corresponding first-level child nodes, wherein the features of the corresponding first-level child nodes include at least one of the following items: The similarity range corresponding to the corresponding first-level child nodes; Whether the corresponding first-level child node has a corresponding first-level child node image.

2. The method according to claim 1, wherein determining the destination node of the input image according to the features of the corresponding first-level child nodes comprises: In response to the corresponding first-level child node having a corresponding first-level child node image, comparing the input image with the corresponding first-level child node image for similarity; as well as According to the similarity comparison result, the input image is transferred to a corresponding second-level child node among a plurality of second-level child nodes, The plurality of second-level child nodes are child nodes of the corresponding first-level child node, and correspond to the plurality of similarity ranges respectively.

3. The method according to claim 1, wherein determining the destination node of the input image according to the features of the corresponding first-level child nodes comprises: In response to the corresponding first-level child node not having a corresponding first-level child node image, the input image is associated with the corresponding first-level child node and saved as the corresponding first-level child node image.

4. The method according to claim 1, wherein determining the destination node of the input image according to the features of the corresponding first-level child nodes comprises: In response to the similarity range corresponding to the corresponding first-level child node having a maximum endpoint value, the input image is associated with the corresponding first-level child node and saved, wherein the input image represents a similar image to the parent node image.

5. The method according to claim 1, wherein the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes a mask image, and the method further comprises: Based on the similarity comparison result, determining that the input image is a similar image to the parent node image; Performing optical proximity effect correction and verification on the time domain image corresponding to the parent node image to obtain a calibrated time domain mask image; as well as The calibrated time-domain mask image is pasted back to the time-domain image corresponding to the input image.

6. The method according to claim 1, wherein the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes an image with defect points, the method further comprising: Based on the similarity comparison result, determining that the input image is a similar image to the parent node image; as well as The same defect repair rule is applied to the time domain image corresponding to the parent node image and the time domain image corresponding to the input image.

7. The method according to claim 1, wherein the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes a sample mask image for training a neural network model.

8. The method according to claim 7, further comprising: Determine a subset of the plurality of first-level child nodes, wherein endpoint values ​​of similarity ranges corresponding to child node elements in the subset are all less than a predetermined similarity threshold; Obtaining a sub-node image at the sub-node element in the subset; as well as The time domain image corresponding to the parent node image and the acquired time domain image corresponding to the child node image are used as sample images for training the neural network model.

9. The method of claim 1, wherein the parent node comprises a root node, the method further comprising: Obtaining a first image from the set of images to be processed; Performing frequency domain transformation on the first image; as well as The first frequency domain image after frequency domain transformation is used as the root node image at the root node.

10. The method according to claim 9, further comprising: Acquire a second image from the set of images to be processed; performing frequency domain transformation on the second image; as well as The second frequency domain image after frequency domain transformation is used as the input image.

11. A system for processing an image, comprising: at least one parent node, a first parent node of the at least one parent node having a corresponding parent node image; as well as a plurality of first-level child nodes of the first parent node, wherein the plurality of first-level child nodes respectively correspond to a plurality of similarity ranges, The system is configured to: Compare the similarity between the input image and the parent node image at the first parent node; According to the similarity comparison result, the input image is transferred to a corresponding first-level child node among the plurality of first-level child nodes of the first parent node; as well as Determining the destination node of the input image according to the features of the corresponding first-level child nodes, The characteristics of the corresponding first-level subnode include at least one of the following items: The similarity range corresponding to the corresponding first-level sub-node; Whether the corresponding first-level child node has a corresponding first-level child node image.

12. The system according to claim 11, wherein determining the destination node of the input image according to the features of the corresponding first-level child nodes comprises: In response to the corresponding first-level child node having a corresponding first-level child node image, comparing the input image with the corresponding first-level child node image for similarity; as well as According to the similarity comparison result, the input image is transferred to a corresponding second-level child node among a plurality of second-level child nodes, The plurality of second-level child nodes are child nodes of the corresponding first-level child node, and correspond to the plurality of similarity ranges respectively.

13. The system according to claim 11, wherein determining the destination node of the input image according to the features of the corresponding first-level child nodes comprises: In response to the corresponding first-level child node not having a corresponding first-level child node image, the input image is associated with the corresponding first-level child node and saved as the corresponding first-level child node image.

14. The system according to claim 11, wherein determining the destination node of the input image according to the features of the corresponding first-level child nodes comprises: In response to the similarity range corresponding to the corresponding first-level child node having a maximum endpoint value, the input image is associated with the corresponding first-level child node and saved, wherein the input image represents a similar image to the parent node image.

15. The system according to claim 11, wherein the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes a mask image, and the system is further configured to: Based on the similarity comparison result, determining that the input image is a similar image to the parent node image; Performing optical proximity effect correction and verification on the time domain image corresponding to the parent node image to obtain a calibrated time domain mask image; and The calibrated time-domain mask image is pasted back to the time-domain image corresponding to the input image.

16. The system according to claim 11, wherein the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes an image with defect points, and the system is further configured to: Based on the similarity comparison result, determining that the input image is a similar image to the parent node image; and The same defect repair rule is applied to the time domain image corresponding to the parent node image and the time domain image corresponding to the input image.

17. The system according to claim 11, wherein the parent node image is a frequency domain image, and the time domain image corresponding to the parent node image includes a sample mask image for training a neural network model.

18. The system according to claim 17, further configured to: Determine a subset of the plurality of first-level child nodes, wherein endpoint values ​​of similarity ranges corresponding to child node elements in the subset are all less than a predetermined similarity threshold; Obtaining a sub-node image at the sub-node element in the subset; and The time domain image corresponding to the parent node image and the acquired time domain image corresponding to the child node image are used as sample images for training the neural network model.

19. The system of claim 11, wherein the first parent node comprises a root node, the system being further configured to: Obtaining a first image from the set of images to be processed; performing frequency domain transformation on the first image; and The first frequency domain image after frequency domain transformation is used as the root node image at the root node.

20. The system according to claim 19, further configured to: Acquire a second image from the set of images to be processed; performing frequency domain transformation on the second image; and The second frequency domain image after frequency domain transformation is used as the input image.

21. An electronic device, comprising: one or more processors; as well as A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to execute the method as described in any one of claims 1 to 10.

22. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 10.

23. A computer program product, comprising a program code, wherein when the program code is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images