Model training method, image processing method, device and medium

Abstract

The disclosed embodiments disclose a model training method, an image processing method, a device and a medium, the method comprising: obtaining at least one first image; obtaining at least one second image corresponding to each first image; obtaining a corresponding filter weight parameter group according to each first image and at least one second image corresponding to each first image; obtaining a filter weight parameter model, taking at least one second image as input and the corresponding filter weight parameter group as output, and training the filter weight parameter model to obtain a trained filter weight parameter model. Based on the trained filter weight parameter model obtained by this scheme, a filter weight parameter group corresponding to the image to be processed can be obtained, and the filter parameters of at least two filters can be adjusted according to the filter weight parameter group, and the image can be processed according to the at least two filters, which can ensure that the display effect of the processed image is more beautiful and improve the user experience.

Description

Technical Field

[0001] The present disclosure relates to the field of machine learning technology, and in particular to a model training method, an image processing method, a device and a medium. Background Art

[0002] In recent years, with the development of e-commerce and the continuous improvement of people's living standards, more and more people are shopping on the corresponding platforms through the Internet. Among them, the images of items displayed on the platform can reflect the various characteristics of the items and play an important role in whether users buy the items. However, since the images of the items are usually uploaded by the merchants themselves, the uploaded images often have problems such as poor quality, unsightly, or images that do not meet the prescribed content. If the images uploaded by the merchants are directly displayed on the corresponding platform, the user experience may be poor when viewing the images, thereby reducing the sales of the corresponding items.

[0003] In the related art, before displaying an image, the image can be processed based on a filter to improve the image quality and make the objects in the processed image more beautiful. However, since the effect presented by the processed image is relatively single, it is impossible to ensure that objects of different categories and objects in different environments are displayed more beautifully in the processed image, resulting in a poor user experience when viewing the processed image. Summary of the invention

[0004] In order to solve the problems in the related art, the embodiments of the present disclosure provide a model training method, an image processing method, a device and a medium.

[0005] In a first aspect, an embodiment of the present disclosure provides a model training method, the method comprising:

[0006] Acquire at least one first image, where the image quality score of the first image is greater than or equal to an image quality score threshold;

[0007] Processing each first image to obtain at least one second image corresponding to each first image, wherein the image quality score of the second image is less than an image quality score threshold;

[0008] According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of the at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain a first image corresponding to the processed second image;

[0009] A filter weight parameter model is obtained, at least one second image corresponding to each first image is used as input, a corresponding filter weight parameter group is used as output, and the filter weight parameter model is trained to obtain a trained filter weight parameter model.

[0010] In one implementation of the present disclosure, the filter weight parameter is used to adjust the filter parameters of the corresponding filter, including:

[0011] The filter weight parameter is used to multiply the filter parameter of the corresponding filter to obtain the adjusted filter parameter;

[0012] The filter weight parameter group includes filter weight parameters corresponding to at least three filters respectively, and the at least three filters include at least two first filters and at least one second filter. The first filter is used to adjust the pixel value of at least one pixel point in the corresponding image, and the second filter is used not to adjust the pixel value of any pixel point in the corresponding image.

[0013] In one implementation of the present disclosure, the method further includes:

[0014] Acquire at least one test image, where the image quality score of the test image is less than an image quality score threshold;

[0015] Input each test image into the trained filter weight parameter model to obtain a filter weight parameter group corresponding to each test image;

[0016] Adjusting filter parameters of the corresponding filter according to the filter weight parameter group corresponding to each test image, and processing each test image by the filter with the adjusted filter parameters to obtain at least one test processed image;

[0017] In response to the image quality score of at least one test processed image being greater than or equal to an image quality score threshold, the trained filter weight parameter model is determined as a target filter weight parameter model.

[0018] In one implementation of the present disclosure, the image quality score is positively correlated with at least one of the clarity of the corresponding image and the color score of the corresponding image;

[0019] And / or, the image quality score is negatively correlated with at least one of the number of target characters, the area occupied by the target characters, the number of target logos, and the area occupied by the target logos in the corresponding image.

[0020] In an implementation of the present disclosure, processing each first image to obtain at least one second image corresponding to each first image includes:

[0021] Processing each first image to obtain a second image corresponding to each first image;

[0022] According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, including:

[0023] According to each first image and a second image corresponding to each first image, a corresponding filter weight parameter group is obtained.

[0024] In one implementation of the present disclosure, each first image is processed, including at least one of adjusting the saturation of each first image, changing the brightness of each first image, blurring at least one area in each first image, and adding noise to each first image.

[0025] In one implementation of the present disclosure, the filter includes a color lookup table.

[0026] In a second aspect, an embodiment of the present disclosure provides an image processing method, the method comprising:

[0027] Acquire at least one sampled image, and input the at least one sampled image into a target filter weight parameter model to acquire a filter weight parameter group corresponding to each sampled image, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters, respectively, and the filter parameters of the at least two filters are different;

[0028] The filter parameters of the corresponding filters are adjusted according to the filter weight parameter group corresponding to each sampled image, and each test image is processed by the filter with the adjusted filter parameters to obtain at least one sampled processed image.

[0029] In a third aspect, an embodiment of the present disclosure provides a model training device, the device comprising:

[0030] A first image acquisition module is configured to acquire at least one first image, wherein the image quality score of the first image is greater than or equal to an image quality score threshold;

[0031] A second image acquisition module is configured to process each first image to acquire at least one second image corresponding to each first image, wherein the image quality score of the second image is less than an image quality score threshold;

[0032] a weight parameter group acquisition module, configured to acquire a corresponding filter weight parameter group according to each first image and at least one second image corresponding to each first image, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of the at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain a first image corresponding to the processed second image;

[0033] The weight parameter model training module is configured to obtain a filter weight parameter model, take at least one second image corresponding to each first image as input, take the corresponding filter weight parameter group as output, train the filter weight parameter model to obtain a trained filter weight parameter model.

[0034] In a fourth aspect, an embodiment of the present disclosure provides an image processing device, the device comprising:

[0035] a sampling image acquisition module, configured to acquire at least one sampling image, and input the at least one sampling image into a target filter weight parameter model to acquire a filter weight parameter group corresponding to each sampling image, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, and the filter parameters of the at least two filters are different;

[0036] The image processing module is configured to adjust the filter parameters of the corresponding filter according to the filter weight parameter group corresponding to each sampled image, and process each test image through the filter with adjusted filter parameters to obtain at least one sampled processed image.

[0037] In a fifth aspect, an embodiment of the present disclosure provides an electronic device, comprising a memory and at least one processor; the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by at least one processor to implement the first aspect, any implementation method of the first aspect, and any method step of the second aspect.

[0038] In a sixth aspect, a computer-readable storage medium is provided in an embodiment of the present disclosure, on which computer instructions are stored. When the computer instructions are executed by a processor, the first aspect, any implementation method of the first aspect, and any method step of the second aspect are implemented.

[0039] In the seventh aspect, a computer program product is provided in an embodiment of the present disclosure, comprising a computer program / instruction, which, when executed by a processor, implements the first aspect, any implementation method of the first aspect, and any method step of the second aspect.

[0040] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0041] In the above embodiment, at least one first image is obtained and each first image is processed to obtain at least one second image corresponding to each first image, wherein the image quality score of the first image is greater than or equal to the image quality score threshold, and the image quality score of the second image is less than the image quality score threshold. Therefore, the first image can be understood as an image with a better display effect of the object and capable of meeting the display effect requirements, and the second image can be understood as an image in which the displayed object and the environment in which the object is located are exactly the same as the first image, but the display effect is poor and cannot meet the display effect requirements. According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain the first image corresponding to the processed second image, that is, to achieve the purpose of improving the image display effect, wherein, since the filter weight parameters are only used to adjust the filter parameters of at least two filters with different filter parameters, therefore, only by adjusting the filter weight parameters themselves, after processing different images including different objects in different environments according to the at least two filters after the filter parameters are adjusted, the display effects of the objects in the processed different images are different, and the display effects are all more beautiful. By obtaining a filter weight parameter model, taking at least one second image corresponding to each first image as input, taking the corresponding filter weight parameter group as output, and training the filter weight parameter model to obtain a trained filter weight parameter model, wherein the trained filter weight parameter model can be understood as a model that has learned the law between the filter weight parameters and the images with poor display effects and unable to meet the requirements. When it is necessary to process different images with poor display effects, unable to meet the display requirements, and including different objects and different environments of the objects, the image to be processed is input into the trained filter weight parameter model to obtain the corresponding filter weight parameter group, and the filter parameters of at least two filters are adjusted according to the filter weight parameters in the filter weight parameter group, and the image to be processed is processed according to the at least two filters after the filter parameters are adjusted, so that the display effects of the objects in different processed images can be relatively beautiful, thereby enriching the display effects of the processed images and making the user experience when viewing the processed images better.

[0042] In the above implementation, considering that in some scenarios, the display effect of the object in the image input to the trained filter weight parameter model may be good and can meet the corresponding display effect requirements, if the trained filter weight parameter model is required to make a large adjustment to the pixel value of the pixel point in the image, it may cause the display effect of the object in the picture after the pixel value of the pixel point is adjusted to be degraded. In order to avoid the occurrence of this situation, the filter weight parameter is limited to be multiplied with the filter parameter of the corresponding filter to obtain the adjusted filter parameter, and the filter weight parameter group is limited to include filter weight parameters corresponding to each of at least three filters, the at least three filters include at least two first filters and at least one second filter, the first filter is used to adjust the pixel value of at least one pixel point in the corresponding image, and the second filter is used not to adjust the pixel value of any pixel point in the corresponding image. Among them, by introducing a second filter for not adjusting the pixel value of any pixel point in the corresponding image, when the display effect of the object in the image input into the trained filter weight parameter model may be better, the filter weight parameters corresponding to the first filter in the filter weight parameter group output by the trained filter weight parameter model can be smaller, while the filter weight parameters corresponding to the second filter can be larger, so as to avoid making a large adjustment to the pixel value of the pixel points in the image. In addition, since the second filter, i.e., the filter for not adjusting the pixel value of any pixel point in the corresponding image, is introduced, in the process of training the filter weight parameter model, while ensuring that the trained filter weight parameter model can learn the rule of not making a large adjustment to the pixel value of the pixel points in the image where the display effect of the object is already better, the training difficulty is reduced and the training efficiency is improved.

[0043] In the above embodiment, at least one test image is obtained, and each test image is input into the trained filter weight parameter model to obtain a filter weight parameter group corresponding to each test image, and the filter parameters of the corresponding filter are adjusted according to the filter weight parameter group corresponding to each test image, and each test image is processed by the filter after the filter parameters are adjusted to obtain at least one test processed image. Wherein, since the image quality score of the test image is less than the image quality score threshold, the test image can be understood as an image in which the display effect of the object is poor and cannot meet the display effect requirements. If the image quality scores of the test processed images obtained after processing are greater than or equal to the image quality score threshold, it can be considered that the test processed image is an image in which the display effect of the object is good and can meet the display effect requirements, that is, it is determined that the trained filter weight parameter model has learned the law between the filter weight parameters and the images with poor display effects that cannot meet the requirements, and the trained filter weight parameter model is determined as the target filter weight parameter model, that is, the trained filter weight parameter model passed the test, thereby achieving the purpose of testing the trained filter weight parameter model.

[0044] In the above implementation, by limiting the image quality score to be positively correlated with at least one of the clarity of the corresponding image and the color score of the corresponding image, it can be ensured that the image with a higher image quality score has a higher clarity and a higher color score, the user has a lower difficulty in viewing the image, can obtain more information, and the color of the object in the image is closer to the color of the object perceived by the human eye in reality, thereby improving the user experience. By limiting the image quality score to be negatively correlated with at least one of the number of target texts, the area occupied by the target texts, the number of target logos, and the area occupied by the target logos in the corresponding image, it can be ensured that the image with a higher image quality score meets at least one of the requirements of the number of texts irrelevant to the image theme in the image is small, the area occupied by texts irrelevant to the image theme in the image is small, the number of badges irrelevant to the image theme in the image is small, and the area occupied by texts irrelevant to the image theme in the image is small, ensuring that the user will not be disturbed by content irrelevant to the image theme when viewing the image, thereby improving the user experience.

[0045] In the above embodiment, by processing each first image to obtain a second image corresponding to each first image, and obtaining the corresponding filter weight parameter group based on each first image and a second image corresponding to each first image, the number of filter weight parameter groups obtained can be reduced, the amount of data processing is reduced, and the speed of obtaining the filter weight parameter groups is accelerated.

[0046] In the above embodiment, by processing each first image to include at least one of adjusting the saturation in each first image, changing the brightness of each first image, blurring at least one area in each first image, and adding noise to each first image, the difficulty of acquiring the second image can be reduced and the efficiency of acquiring the second image can be improved.

[0047] In the above implementation, by limiting the filter to include a color lookup table, the amount of calculation when processing the corresponding image based on the filter can be reduced, thereby improving the processing efficiency.

[0048] In the above implementation, at least one sampled image is obtained and input into the target filter weight parameter model to obtain a filter weight parameter group corresponding to each sampled image, including filter weight parameters corresponding to at least two filters respectively, wherein the filter parameters of at least two filters are different, and the filter parameters of the corresponding filter are adjusted according to the filter weight parameter group corresponding to each sampled image, and each test image is processed by the filter after the filter parameters are adjusted to obtain at least one sampled processed image. Among them, the trained filter weight parameter model can be understood as a model that has learned the law between the filter weight parameters and images with poor display effects that cannot meet the requirements. When it is necessary to process different images with poor display effects, which cannot meet the display requirements, and which include different objects and are in different environments, the images to be processed are input into the trained filter weight parameter model to obtain the corresponding filter weight parameter group, and the filter parameters of at least two filters are adjusted according to the filter weight parameters in the filter weight parameter group, and the images to be processed are processed according to the at least two filters after the filter parameters are adjusted, so that the display effects of the objects in the different processed images can be more beautiful. Moreover, since the filter weight parameters are only used to adjust the filter parameters of at least two filters with different filter parameters, only by adjusting the filter weight parameters themselves, different images including different objects in different environments can be processed according to the at least two filters after the filter parameters are adjusted, and the display effects of the objects in the processed different images are different, and the display effects are all more beautiful. Therefore, the above scheme can enrich the display effect of the processed images, and ensure that the display effect of the processed images is more beautiful, so that the user experience when viewing the processed images is better.

[0049] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0051] Figure 1 A flow chart showing a model training method according to an embodiment of the present disclosure;

[0052] Figure 2 A flowchart showing an image processing method according to an embodiment of the present disclosure is shown;

[0053] Figure 3 A structural block diagram of a model training device according to an embodiment of the present disclosure is shown;

[0054] Figure 4 A structural block diagram of an image processing device according to an embodiment of the present disclosure is shown;

[0055] Figure 5 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown;

[0056] Figure 6 It is a schematic diagram of the structure of a computer system suitable for implementing the method according to the embodiment of the present disclosure. DETAILED DESCRIPTION

[0057] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0058] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate the presence of labels, numbers, steps, behaviors, components, parts, or combinations thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other labels, numbers, steps, behaviors, components, parts, or combinations thereof exist or are added.

[0059] It should also be noted that, in the absence of conflict, the embodiments and labels in the embodiments of the present disclosure can be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0060] In order to process the image of the object, the inventors of the present disclosure considered the following solution.

[0061] The image can be processed based on the filter obtained in advance to improve the image quality, so that the objects in the processed image are more beautiful and the display effect is better. In order to ensure that different types of objects and objects in different environments have better display effects in the processed image, multiple filters can be obtained in advance, and different filters can be used for images of different types of objects and objects in different environments.

[0062] Exemplarily, when the images to be processed are multiple images of take-out items, and among the multiple images, the take-out items include stir-fried dishes, noodles, buns, etc., and the environment in which the take-out items are located includes a take-out container, on a plate, etc., a filter for processing the image of the stir-fried dishes in the take-out container, a filter for processing the image of the noodles in the take-out container, a filter for processing the image of the buns in the take-out container, a filter for processing the image of the stir-fried dishes in the plate, a filter for processing the image of the noodles in the plate, and a filter for processing the image of the buns in the plate can be obtained, and different images are processed based on different filters respectively to ensure that different types of items and items in different environments have better display effects in the processed images.

[0063] Disadvantages of this solution: In recent years, the types of objects have become more and more abundant, and the types of environments in which objects are located in different images have become more and more diverse. Considering the high cost of obtaining filters, the number of filters pre-acquired in the above solution is small, resulting in the display effect of objects in the images processed according to the pre-acquired filters being relatively simple. Therefore, it is impossible to ensure that objects of different categories and objects located in different environments are displayed beautifully in the processed images, resulting in a poor user experience when users view the processed images.

[0064] Taking into account the shortcomings of the above schemes, the inventors of the present disclosure have proposed a new scheme: by acquiring at least one first image, processing each first image to obtain at least one second image corresponding to each first image, wherein the image quality score of the first image is greater than or equal to the image quality score threshold, and the image quality score of the second image is less than the image quality score threshold. Therefore, the first image can be understood as an image with a better display effect of the object and able to meet the display effect requirements, and the second image can be understood as an image in which the displayed object and the environment in which the object is located are exactly the same as the first image, but the display effect is poor and cannot meet the display effect requirements. According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain the first image corresponding to the processed second image, that is, to achieve the purpose of improving the image display effect, wherein, since the filter weight parameters are only used to adjust the filter parameters of at least two filters with different filter parameters, therefore, only by adjusting the filter weight parameters themselves, after processing different images including different objects in different environments according to the at least two filters after the filter parameters are adjusted, the display effects of the objects in the processed different images are different, and the display effects are all more beautiful. By obtaining a filter weight parameter model, taking at least one second image corresponding to each first image as input, taking the corresponding filter weight parameter group as output, and training the filter weight parameter model to obtain a trained filter weight parameter model, wherein the trained filter weight parameter model can be understood as a model that has learned the law between the filter weight parameters and the images with poor display effects and unable to meet the requirements. When it is necessary to process different images with poor display effects, unable to meet the display requirements, and including different objects and different environments of the objects, the image to be processed is input into the trained filter weight parameter model to obtain the corresponding filter weight parameter group, and the filter parameters of at least two filters are adjusted according to the filter weight parameters in the filter weight parameter group, and the image to be processed is processed according to the at least two filters after the filter parameters are adjusted, so that the display effects of the objects in different processed images can be relatively beautiful, thereby enriching the display effects of the processed images and making the user experience when viewing the processed images better.

[0065] In order to solve the above problems, the present disclosure proposes a model training method, an image processing method, a device and a medium.

[0066] Figure 1 FIG. 1 is a flow chart of a model training method according to an embodiment of the present disclosure. Figure 1As shown, the model training method includes steps S101-S104.

[0067] In step S101, at least one first image is acquired.

[0068] wherein the image quality score of the first image is greater than or equal to an image quality score threshold;

[0069] In step S102, each first image is processed to obtain at least one second image corresponding to each first image.

[0070] The image quality score of the second image is less than an image quality score threshold.

[0071] In step S103, a corresponding filter weight parameter group is obtained according to each first image and at least one second image corresponding to each first image.

[0072] Among them, the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain a first image corresponding to the processed second image.

[0073] In step S104, a filter weight parameter model is obtained, at least one second image corresponding to each first image is used as input, and the corresponding filter weight parameter group is used as output, and the filter weight parameter model is trained to obtain a trained filter weight parameter model.

[0074] In one embodiment of the present disclosure, acquiring at least one first image may be reading at least one first image stored in advance, or may be acquiring at least one first image from other devices or systems.

[0075] In one embodiment of the present disclosure, the image quality score can be understood as an indicator of the quality of the display effect of the corresponding image on the object in the image. For example, when an image is clearer, the content irrelevant to the subject of the image is less, and the color of the image is closer to the real color of the object in the image, the image quality score of the image is higher, and vice versa.

[0076] The image quality score of the first image can be obtained by inputting the image data of the first image or the feature data of the first image into calculation based on a corresponding algorithm; or the image quality score of the first image can be obtained by inputting a pre-trained image quality scoring model into the first image to output the image quality score of the first image.

[0077] In one embodiment of the present disclosure, obtaining at least one second image corresponding to each first image can be understood as processing each first image based on a pre-acquired quality score reduction algorithm to obtain at least one second image corresponding to each first image; it can also be understood as obtaining a pre-trained quality score reduction model and inputting each first image into the quality score reduction model to obtain output at least one second image corresponding to each first image.

[0078] In one embodiment of the present disclosure, the second image can be understood as that the object displayed therein is the same as the object displayed in the first image, and the environment in which the object is located in the second image is also the same as the environment in which the object is located in the first image, but the display effect of the object in the first image is better than the display effect of the object in the second image. Although the content obtained by the user when viewing the second image is the same as the content obtained by viewing the first image, the user experience of the user when viewing the second image is worse than the user experience of the user when viewing the first image.

[0079] In one embodiment of the present disclosure, a filter can be understood as being used to adjust the pixel value of a corresponding pixel point in an image to adjust the display effect of the image. A filter parameter can be understood as being used to indicate the amplitude by which a filter adjusts the pixel value of a corresponding pixel point in an image. A filter weight parameter can be understood as being used to adjust the filter parameter of a corresponding filter. Adjusting the filter parameter of a corresponding filter according to a filter weight parameter can be understood as substituting the filter weight parameter and the filter parameter of the corresponding filter into a calculation based on a pre-acquired filter parameter adjustment algorithm to obtain the adjusted filter parameter. The present disclosure does not limit the specific implementation method of adjusting the filter parameter of a corresponding filter according to the filter weight parameter. Exemplarily, when a filter weight parameter group includes filter weight parameters corresponding to 9 filters with different filter parameters, the value ranges of the 9 filter weight parameters in the filter weight parameter group are all greater than 0 and less than 1, and the sum of the 9 filter weight parameters is equal to 1. By multiplying the filter weight parameter with the filter parameter of the corresponding filter, the adjusted filter parameter of the filter can be obtained.

[0080] In one embodiment of the present disclosure, a corresponding filter weight parameter group is obtained according to each first image and at least one second image corresponding to each first image. This can be understood as, based on a pre-acquired algorithm, bringing the image data of each first image and the image data of at least one second image corresponding to each first image into calculation to obtain a filter weight parameter group corresponding to each first image and at least one second image corresponding to each first image; or, in order to obtain a pre-trained filter weight parameter group model, each first image and at least one second image corresponding to each first image are input into the filter weight parameter group model to obtain a filter weight parameter group corresponding to each first image and at least one second image corresponding to each first image.

[0081] Exemplarily, a filter weight parameter group can be randomly generated, and filter parameters of at least two filters can be adjusted according to the randomly generated filter weight parameter group, and at least one second image can be processed according to the at least two filters after the filter parameters are adjusted. When the similarity between a first image corresponding to the at least one second image and the at least one processed second image is greater than or equal to a similarity threshold, the randomly generated filter weight parameter group can be determined as the filter weight parameter group corresponding to the first image and at least one second image corresponding to the first image.

[0082] In one embodiment of the present disclosure, the filter weight parameter model can be a neural network (NN) model, a convolutional neural network (CNN) model or a long short-term memory network (LSTM) model, etc. The present disclosure does not limit the specific implementation method of the filter weight parameter model.

[0083] In the above embodiment, at least one first image is obtained and each first image is processed to obtain at least one second image corresponding to each first image, wherein the image quality score of the first image is greater than or equal to the image quality score threshold, and the image quality score of the second image is less than the image quality score threshold. Therefore, the first image can be understood as an image with a better display effect of the object and capable of meeting the display effect requirements, and the second image can be understood as an image in which the displayed object and the environment in which the object is located are exactly the same as the first image, but the display effect is poor and cannot meet the display effect requirements. According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain the first image corresponding to the processed second image, that is, to achieve the purpose of improving the image display effect, wherein, since the filter weight parameters are only used to adjust the filter parameters of at least two filters with different filter parameters, therefore, only by adjusting the filter weight parameters themselves, after processing different images including different objects in different environments according to the at least two filters after the filter parameters are adjusted, the display effects of the objects in the processed different images are different, and the display effects are all more beautiful. By obtaining a filter weight parameter model, taking at least one second image corresponding to each first image as input, taking the corresponding filter weight parameter group as output, and training the filter weight parameter model to obtain a trained filter weight parameter model, wherein the trained filter weight parameter model can be understood as a model that has learned the law between the filter weight parameters and the images with poor display effects and unable to meet the requirements. When it is necessary to process different images with poor display effects, unable to meet the display requirements, and including different objects and different environments of the objects, the image to be processed is input into the trained filter weight parameter model to obtain the corresponding filter weight parameter group, and the filter parameters of at least two filters are adjusted according to the filter weight parameters in the filter weight parameter group, and the image to be processed is processed according to the at least two filters after the filter parameters are adjusted, so that the display effects of the objects in different processed images can be relatively beautiful, thereby enriching the display effects of the processed images and making the user experience when viewing the processed images better.

[0084] In one implementation of the present disclosure, the filter weight parameter is used to adjust the filter parameters of the corresponding filter, including:

[0085] The filter weight parameter is used to multiply the filter parameter of the corresponding filter to obtain the adjusted filter parameter;

[0086] The filter weight parameter group includes filter weight parameters corresponding to at least three filters respectively, and the at least three filters include at least two first filters and at least one second filter. The first filter is used to adjust the pixel value of at least one pixel point in the corresponding image, and the second filter is used not to adjust the pixel value of any pixel point in the corresponding image.

[0087] In one embodiment of the present disclosure, the second filter is used not to adjust the pixel value of any pixel point in the corresponding image. It can be understood that the filter parameters of the second filter are all 0. Even if the filter weight parameters corresponding to the second filter are multiplied by the filter parameters of the second filter to obtain the adjusted filter parameters of the second filter, the second filter after the filter parameters are adjusted will not adjust the pixel value of any pixel point in the corresponding image.

[0088] In the above implementation, considering that in some scenarios, the display effect of the object in the image input to the trained filter weight parameter model may be good and can meet the corresponding display effect requirements, if the trained filter weight parameter model is required to make a large adjustment to the pixel value of the pixel point in the image, it may cause the display effect of the object in the picture after the pixel value of the pixel point is adjusted to be degraded. In order to avoid the occurrence of this situation, the filter weight parameter is limited to be multiplied with the filter parameter of the corresponding filter to obtain the adjusted filter parameter, and the filter weight parameter group is limited to include filter weight parameters corresponding to each of at least three filters, the at least three filters include at least two first filters and at least one second filter, the first filter is used to adjust the pixel value of at least one pixel point in the corresponding image, and the second filter is used not to adjust the pixel value of any pixel point in the corresponding image. Among them, by introducing a second filter for not adjusting the pixel value of any pixel point in the corresponding image, when the display effect of the object in the image input into the trained filter weight parameter model may be better, the filter weight parameters corresponding to the first filter in the filter weight parameter group output by the trained filter weight parameter model can be smaller, while the filter weight parameters corresponding to the second filter can be larger, so as to avoid making a large adjustment to the pixel value of the pixel points in the image. In addition, since the second filter, i.e., the filter for not adjusting the pixel value of any pixel point in the corresponding image, is introduced, in the process of training the filter weight parameter model, while ensuring that the trained filter weight parameter model can learn the rule of not making a large adjustment to the pixel value of the pixel points in the image where the display effect of the object is already better, the training difficulty is reduced and the training efficiency is improved.

[0089] In one implementation of the present disclosure, the method further includes the following steps:

[0090] Acquire at least one test image, where the image quality score of the test image is less than an image quality score threshold;

[0091] Input each test image into the trained filter weight parameter model to obtain a filter weight parameter group corresponding to each test image;

[0092] Adjusting filter parameters of the corresponding filter according to the filter weight parameter group corresponding to each test image, and processing each test image by the filter with the adjusted filter parameters to obtain at least one test processed image;

[0093] In response to the image quality score of at least one test processed image being greater than or equal to an image quality score threshold, the trained filter weight parameter model is determined as a target filter weight parameter model.

[0094] In the above embodiment, at least one test image is obtained, and each test image is input into the trained filter weight parameter model to obtain a filter weight parameter group corresponding to each test image, and the filter parameters of the corresponding filter are adjusted according to the filter weight parameter group corresponding to each test image, and each test image is processed by the filter after the filter parameters are adjusted to obtain at least one test processed image. Wherein, since the image quality score of the test image is less than the image quality score threshold, the test image can be understood as an image in which the display effect of the object is poor and cannot meet the display effect requirements. If the image quality scores of the test processed images obtained after processing are greater than or equal to the image quality score threshold, it can be considered that the test processed image is an image in which the display effect of the object is good and can meet the display effect requirements, that is, it is determined that the trained filter weight parameter model has learned the law between the filter weight parameters and the images with poor display effects that cannot meet the requirements, and the trained filter weight parameter model is determined as the target filter weight parameter model, that is, the trained filter weight parameter model passed the test, thereby achieving the purpose of testing the trained filter weight parameter model.

[0095] In one implementation of the present disclosure, the image quality score is positively correlated with at least one of the clarity of the corresponding image and the color score of the corresponding image;

[0096] And / or, the image quality score is negatively correlated with at least one of the number of target characters, the area occupied by the target characters, the number of target logos, and the area occupied by the target logos in the corresponding image.

[0097] In one embodiment of the present disclosure, the clarity of an image can be understood as indicating the resolution of the entire image, or as indicating the resolution of at least one region in the image, or as indicating the richness of object detail features of an object in the image. Exemplarily, the higher the clarity of an image, the higher the resolution of the image, and the richer the object detail features of the object in the image.

[0098] The color score of an image can be understood as an indicator of the difference between the color of the corresponding position of an object in the image and the color of the corresponding position of the object perceived by the human eye in reality. The higher the color score, the smaller the difference between the two, that is, the color of the object in the image is closer to the color of the object perceived by the human eye in reality.

[0099] The target text in the corresponding image can be understood as other text in the corresponding image except the text related to the subject of the corresponding image. For example, when the object displayed in the corresponding image includes a hamburger, and the corresponding image also includes text such as the name of a fitness center and a telephone number, then this part of the text belongs to the target text in the corresponding image.

[0100] The target logo in the corresponding image can be understood as other badges in the corresponding image except for the designated badge. The badge can be a trademark or other indicative patterns.

[0101] In the above implementation, by limiting the image quality score to be positively correlated with at least one of the clarity of the corresponding image and the color score of the corresponding image, it can be ensured that the image with a higher image quality score has a higher clarity and a higher color score, the user has a lower difficulty in viewing the image, can obtain more information, and the color of the object in the image is closer to the color of the object perceived by the human eye in reality, thereby improving the user experience. By limiting the image quality score to be negatively correlated with at least one of the number of target texts, the area occupied by the target texts, the number of target logos, and the area occupied by the target logos in the corresponding image, it can be ensured that the image with a higher image quality score meets at least one of the requirements of the number of texts irrelevant to the image theme in the image is small, the area occupied by texts irrelevant to the image theme in the image is small, the number of badges irrelevant to the image theme in the image is small, and the area occupied by texts irrelevant to the image theme in the image is small, ensuring that the user will not be disturbed by content irrelevant to the image theme when viewing the image, thereby improving the user experience.

[0102] In one implementation of the present disclosure, in step S102, processing each first image to obtain at least one second image corresponding to each first image can be implemented by the following steps:

[0103] Processing each first image to obtain a second image corresponding to each first image;

[0104] In step S102, according to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, which can be achieved by the following steps:

[0105] According to each first image and a second image corresponding to each first image, a corresponding filter weight parameter group is obtained.

[0106] In the above embodiment, by processing each first image to obtain a second image corresponding to each first image, and obtaining the corresponding filter weight parameter group based on each first image and a second image corresponding to each first image, the number of filter weight parameter groups obtained can be reduced, the amount of data processing is reduced, and the speed of obtaining the filter weight parameter groups is accelerated.

[0107] In one implementation of the present disclosure, each first image is processed, including at least one of adjusting the saturation of each first image, changing the brightness of each first image, blurring at least one area in each first image, and adding noise to each first image.

[0108] In one embodiment of the present disclosure, the saturation in an image can be adjusted by substituting the image data of the corresponding image into a saturation adjustment algorithm for calculation to obtain an image after the saturation adjustment; or the corresponding image can be input into a pre-acquired saturation adjustment model to obtain an output image after the saturation adjustment.

[0109] In one embodiment of the present disclosure, the brightness in an image can be adjusted by substituting image data of a corresponding image into a brightness adjustment algorithm for calculation to obtain an image after brightness adjustment; or the corresponding image can be input into a pre-acquired brightness adjustment model to obtain an output image after brightness adjustment.

[0110] In one embodiment of the present disclosure, blurring of an image can be performed by substituting image data of a corresponding image into a blurring processing algorithm for calculation to obtain a blurred image; or the corresponding image can be input into a blurring processing model for calculation to obtain an output blurred image.

[0111] In one embodiment of the present disclosure, noise is added to an image by substituting image data of a corresponding image into a noise adding algorithm for calculation to obtain an image after noise is added; or the corresponding image is input into a noise adding model obtained in advance to obtain an output image after noise is added.

[0112] In the above embodiment, by processing each first image to include at least one of adjusting the saturation in each first image, changing the brightness of each first image, blurring at least one area in each first image, and adding noise to each first image, the difficulty of acquiring the second image can be reduced and the efficiency of acquiring the second image can be improved.

[0113] In one implementation of the present disclosure, the filter includes a color lookup table.

[0114] In the above implementation, by limiting the filter to include a color lookup table (LUT), the amount of calculation when processing the corresponding image based on the filter can be reduced, thereby improving the processing efficiency.

[0115] Figure 2 FIG. 1 is a flowchart showing an image processing method according to an embodiment of the present disclosure. Figure 2 As shown, the image processing method includes steps S201-S202.

[0116] In step S201, at least one sampled image is obtained, and at least one sampled image is input into a target filter weight parameter model to obtain a filter weight parameter group corresponding to each sampled image.

[0117] The filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, and the filter parameters of the at least two filters are different.

[0118] In step S202, filter parameters of the corresponding filters are adjusted according to the filter weight parameter group corresponding to each sampled image, and each test image is processed by the filter with the adjusted filter parameters to obtain at least one sampled processed image.

[0119] In one embodiment of the present disclosure, obtaining at least one sampling image may be reading at least one sampling image stored in advance, or may be obtaining at least one sampling image from other devices or systems.

[0120] In one embodiment of the present disclosure, a filter can be understood as being used to adjust the pixel value of a corresponding pixel point in an image to adjust the display effect of the image. A filter parameter can be understood as being used to indicate the amplitude by which a filter adjusts the pixel value of a corresponding pixel point in an image. A filter weight parameter can be understood as being used to adjust the filter parameter of a corresponding filter. Adjusting the filter parameter of a corresponding filter according to a filter weight parameter can be understood as substituting the filter weight parameter and the filter parameter of the corresponding filter into a calculation based on a pre-acquired filter parameter adjustment algorithm to obtain the adjusted filter parameter. The present disclosure does not limit the specific implementation method of adjusting the filter parameter of a corresponding filter according to the filter weight parameter. Exemplarily, when a filter weight parameter group includes filter weight parameters corresponding to 9 filters with different filter parameters, the value ranges of the 9 filter weight parameters in the filter weight parameter group are all greater than 0 and less than 1, and the sum of the 9 filter weight parameters is equal to 1. By multiplying the filter weight parameter with the filter parameter of the corresponding filter, the adjusted filter parameter of the filter can be obtained.

[0121] In one embodiment of the present disclosure, the filter weight parameter model can be a neural network (NN) model, a convolutional neural network (CNN) model or a long short-term memory network (LSTM) model, etc. The present disclosure does not limit the specific implementation method of the filter weight parameter model.

[0122] In the above implementation, at least one sampled image is obtained and input into the target filter weight parameter model to obtain a filter weight parameter group corresponding to each sampled image, including filter weight parameters corresponding to at least two filters respectively, wherein the filter parameters of at least two filters are different, and the filter parameters of the corresponding filter are adjusted according to the filter weight parameter group corresponding to each sampled image, and each test image is processed by the filter after the filter parameters are adjusted to obtain at least one sampled processed image. Among them, the trained filter weight parameter model can be understood as a model that has learned the law between the filter weight parameters and images with poor display effects that cannot meet the requirements. When it is necessary to process different images with poor display effects, which cannot meet the display requirements, and which include different objects and are in different environments, the images to be processed are input into the trained filter weight parameter model to obtain the corresponding filter weight parameter group, and the filter parameters of at least two filters are adjusted according to the filter weight parameters in the filter weight parameter group, and the images to be processed are processed according to the at least two filters after the filter parameters are adjusted, so that the display effects of the objects in the different processed images can be more beautiful. Moreover, since the filter weight parameters are only used to adjust the filter parameters of at least two filters with different filter parameters, only by adjusting the filter weight parameters themselves, different images including different objects in different environments can be processed according to the at least two filters after the filter parameters are adjusted, and the display effects of the objects in the processed different images are different, and the display effects are all more beautiful. Therefore, the above scheme can enrich the display effect of the processed images, and ensure that the display effect of the processed images is more beautiful, so that the user experience when viewing the processed images is better.

[0123] The following reference Figure 3 A model training device according to an embodiment of the present disclosure is described. Figure 3 A structural block diagram of a model training device 100 according to an embodiment of the present disclosure is shown.

[0124] like Figure 3 As shown, the model training device 100 includes:

[0125] The first image acquisition module 101 is configured to acquire at least one first image, wherein the image quality score of the first image is greater than or equal to an image quality score threshold;

[0126] The second image acquisition module 102 is configured to process each first image to acquire at least one second image corresponding to each first image, wherein the image quality score of the second image is less than an image quality score threshold;

[0127] The weight parameter group acquisition module 103 is configured to acquire a corresponding filter weight parameter group according to each first image and at least one second image corresponding to each first image, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of the at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain a first image corresponding to the processed second image;

[0128] The weight parameter model training module 104 is configured to obtain a filter weight parameter model, take at least one second image corresponding to each first image as input, take the corresponding filter weight parameter group as output, train the filter weight parameter model to obtain a trained filter weight parameter model.

[0129] In the above embodiment, at least one first image is obtained and each first image is processed to obtain at least one second image corresponding to each first image, wherein the image quality score of the first image is greater than or equal to the image quality score threshold, and the image quality score of the second image is less than the image quality score threshold. Therefore, the first image can be understood as an image with a better display effect of the object and capable of meeting the display effect requirements, and the second image can be understood as an image in which the displayed object and the environment in which the object is located are exactly the same as the first image, but the display effect is poor and cannot meet the display effect requirements. According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain the first image corresponding to the processed second image, that is, to achieve the purpose of improving the image display effect, wherein, since the filter weight parameters are only used to adjust the filter parameters of at least two filters with different filter parameters, therefore, only by adjusting the filter weight parameters themselves, after processing different images including different objects in different environments according to the at least two filters after the filter parameters are adjusted, the display effects of the objects in the processed different images are different, and the display effects are all more beautiful. By obtaining a filter weight parameter model, taking at least one second image corresponding to each first image as input, taking the corresponding filter weight parameter group as output, and training the filter weight parameter model to obtain a trained filter weight parameter model, wherein the trained filter weight parameter model can be understood as a model that has learned the law between the filter weight parameters and the images with poor display effects and unable to meet the requirements. When it is necessary to process different images with poor display effects, unable to meet the display requirements, and including different objects and different environments of the objects, the image to be processed is input into the trained filter weight parameter model to obtain the corresponding filter weight parameter group, and the filter parameters of at least two filters are adjusted according to the filter weight parameters in the filter weight parameter group, and the image to be processed is processed according to the at least two filters after the filter parameters are adjusted, so that the display effects of the objects in different processed images can be relatively beautiful, thereby enriching the display effects of the processed images and making the user experience when viewing the processed images better.

[0130] Those skilled in the art will understand that referring to Figure 3 The described technical solution can be combined with any of the embodiments described above, so as to have the technical effect achieved by any of the embodiments described above. The specific contents can refer to the description of the above embodiments, and the specific contents will not be repeated here.

[0131] The following reference Figure 4 An image processing apparatus according to an embodiment of the present disclosure is described. Figure 4 A structural block diagram of an image processing device 200 according to an embodiment of the present disclosure is shown.

[0132] like Figure 4 As shown, the image processing device 200 includes:

[0133] The sampling image acquisition module 201 is configured to acquire at least one sampling image, and input the at least one sampling image into the target filter weight parameter model to acquire a filter weight parameter group corresponding to each sampling image, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, and the filter parameters of the at least two filters are different;

[0134] The image processing module 202 is configured to adjust the filter parameters of the corresponding filter according to the filter weight parameter group corresponding to each sampled image, and process each test image through the filter with the adjusted filter parameters to obtain at least one sampled processed image.

[0135] In the above implementation, at least one sampled image is obtained and input into the target filter weight parameter model to obtain a filter weight parameter group corresponding to each sampled image, including filter weight parameters corresponding to at least two filters respectively, wherein the filter parameters of at least two filters are different, and the filter parameters of the corresponding filter are adjusted according to the filter weight parameter group corresponding to each sampled image, and each test image is processed by the filter after the filter parameters are adjusted to obtain at least one sampled processed image. Among them, the trained filter weight parameter model can be understood as a model that has learned the law between the filter weight parameters and images with poor display effects that cannot meet the requirements. When it is necessary to process different images with poor display effects, which cannot meet the display requirements, and which include different objects and are in different environments, the images to be processed are input into the trained filter weight parameter model to obtain the corresponding filter weight parameter group, and the filter parameters of at least two filters are adjusted according to the filter weight parameters in the filter weight parameter group, and the images to be processed are processed according to the at least two filters after the filter parameters are adjusted, so that the display effects of the objects in the different processed images can be more beautiful. Moreover, since the filter weight parameters are only used to adjust the filter parameters of at least two filters with different filter parameters, only by adjusting the filter weight parameters themselves, different images including different objects in different environments can be processed according to the at least two filters after the filter parameters are adjusted, and the display effects of the objects in the processed different images are different, and the display effects are all more beautiful. Therefore, the above scheme can enrich the display effect of the processed images, and ensure that the display effect of the processed images is more beautiful, so that the user experience when viewing the processed images is better.

[0136] Those skilled in the art will understand that referring to Figure 4 The described technical solution can be combined with any of the embodiments described above, so as to have the technical effect achieved by any of the embodiments described above. The specific contents can refer to the description of the above embodiments, and the specific contents will not be repeated here.

[0137] Figure 5 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0138] The present disclosure also provides an electronic device, such as Figure 5 As shown, it includes at least one processor 301; and a memory 302 that is communicatively connected to the at least one processor 301; wherein the memory 302 stores instructions that can be executed by the at least one processor 301, and the instructions are executed by the at least one processor 301 to implement the following steps:

[0139] In a first aspect, an embodiment of the present disclosure provides a model training method, the method comprising:

[0140] Acquire at least one first image, where the image quality score of the first image is greater than or equal to an image quality score threshold;

[0141] Processing each first image to obtain at least one second image corresponding to each first image, wherein the image quality score of the second image is less than an image quality score threshold;

[0142] According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of the at least two filters are different, the filter weight parameters are used to adjust the filter parameters of the corresponding filters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain a first image corresponding to the processed second image;

[0143] A filter weight parameter model is obtained, at least one second image corresponding to each first image is used as input, a corresponding filter weight parameter group is used as output, and the filter weight parameter model is trained to obtain a trained filter weight parameter model.

[0144] In one implementation of the present disclosure, the filter weight parameter is used to adjust the filter parameters of the corresponding filter, including:

[0145] The filter weight parameter is used to multiply the filter parameter of the corresponding filter to obtain the adjusted filter parameter;

[0146] The filter weight parameter group includes filter weight parameters corresponding to at least three filters respectively, and the at least three filters include at least two first filters and at least one second filter. The first filter is used to adjust the pixel value of at least one pixel point in the corresponding image, and the second filter is used not to adjust the pixel value of any pixel point in the corresponding image.

[0147] In one implementation of the present disclosure, the method further includes:

[0148] Acquire at least one test image, where the image quality score of the test image is less than an image quality score threshold;

[0149] Input each test image into the trained filter weight parameter model to obtain a filter weight parameter group corresponding to each test image;

[0150] Adjusting filter parameters of the corresponding filter according to the filter weight parameter group corresponding to each test image, and processing each test image by the filter with the adjusted filter parameters to obtain at least one test processed image;

[0151] In response to the image quality score of at least one test processed image being greater than or equal to an image quality score threshold, the trained filter weight parameter model is determined as a target filter weight parameter model.

[0152] In one implementation of the present disclosure, the image quality score is positively correlated with at least one of the clarity of the corresponding image and the color score of the corresponding image;

[0153] And / or, the image quality score is negatively correlated with at least one of the number of target characters, the area occupied by the target characters, the number of target logos, and the area occupied by the target logos in the corresponding image.

[0154] In an implementation of the present disclosure, processing each first image to obtain at least one second image corresponding to each first image includes:

[0155] Processing each first image to obtain a second image corresponding to each first image;

[0156] According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, including:

[0157] According to each first image and a second image corresponding to each first image, a corresponding filter weight parameter group is obtained.

[0158] In one implementation of the present disclosure, each first image is processed, including at least one of adjusting the saturation of each first image, changing the brightness of each first image, blurring at least one area in each first image, and adding noise to each first image.

[0159] In one implementation of the present disclosure, the filter includes a color lookup table.

[0160] In a second aspect, an embodiment of the present disclosure provides an image processing method, the method comprising:

[0161] Acquire at least one sampled image, and input the at least one sampled image into a target filter weight parameter model to acquire a filter weight parameter group corresponding to each sampled image, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters, respectively, and the filter parameters of the at least two filters are different;

[0162] The filter parameters of the corresponding filters are adjusted according to the filter weight parameter group corresponding to each sampled image, and each test image is processed by the filter with the adjusted filter parameters to obtain at least one sampled processed image.

[0163] like Figure 6 As shown, the computer system 400 includes a processing unit 401, which can perform various processes in the embodiments shown in the above figures according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 to the random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the system 400 are also stored. The CPU 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0164] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. The drive 410 is also connected to the I / O interface 405 as needed. Removable media 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., are installed on the drive 410 as needed, so that the computer program read therefrom is installed into the storage section 408 as needed. Among them, the processing unit 401 can be implemented as a processing unit such as a CPU, a GPU, a TPU, an FPGA, an NPU, etc.

[0165] In particular, according to an embodiment of the present disclosure, the method described above with reference to the accompanying drawings can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program tangibly contained on a readable medium thereof, and the computer program includes a program code for executing the method in the accompanying drawings. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 409, and / or installed from a removable medium 411. For example, an embodiment of the present disclosure includes a readable storage medium on which computer instructions are stored, and when the computer instructions are executed by a processor, the program code for executing the method in the accompanying drawings is implemented.

[0166] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0167] The units or modules involved in the embodiments described in the present disclosure may be implemented by software or hardware. The units or modules described may also be set in a processor, and the names of these units or modules do not constitute limitations on the units or modules themselves in some cases.

[0168] As another aspect, the present disclosure further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the node described in the above embodiment; or a computer-readable storage medium that exists independently and is not installed in a device. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the method described in the present disclosure.

[0169] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other.

Claims

1. A model training method, wherein: The method comprises: Acquire at least one first image, where the image quality score of the first image is greater than or equal to an image quality score threshold; Processing each first image to obtain at least one second image corresponding to each first image, wherein the image quality score of the second image is less than the image quality score threshold; According to each first image and at least one second image corresponding to each first image, a corresponding filter weight parameter group is obtained, the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, the filter parameters of the at least two filters are different, and the filter weight parameters are used to adjust the filter parameters of the corresponding filters, including: the filter weight parameters are used to multiply the filter parameters of the corresponding filters to obtain adjusted filter parameters, and the at least two filters after the filter parameters are adjusted are used to process the second image to obtain a first image corresponding to the processed second image; the at least two filters include a first filter and a second filter, the first filter is used to adjust the pixel value of at least one pixel point in the corresponding image, and the second filter is used not to adjust the pixel value of any pixel point in the corresponding image; A filter weight parameter model is obtained, at least one second image corresponding to each first image is used as input, a corresponding filter weight parameter group is used as output, and the filter weight parameter model is trained to obtain a trained filter weight parameter model.

2. The model training method according to claim 1, wherein: The filter weight parameter group includes filter weight parameters corresponding to at least three filters respectively, and the at least three filters include at least two first filters and at least one second filter.

3. The model training method according to claim 1, wherein: The method further comprises: Acquire at least one test image, wherein the image quality score of the test image is less than the image quality score threshold; Inputting each test image into the trained filter weight parameter model to obtain a filter weight parameter group corresponding to each test image; Adjusting filter parameters of corresponding filters according to the filter weight parameter group corresponding to each test image, and processing each test image by the filter with the adjusted filter parameters to obtain at least one test processed image; In response to the image quality score of the at least one test processed image being greater than or equal to the image quality score threshold, the trained filter weight parameter model is determined as a target filter weight parameter model.

4. The model training method according to any one of claims 1 to 3, wherein: The image quality score is positively correlated with at least one of the clarity of the corresponding image and the color score of the corresponding image; And / or, the image quality score is negatively correlated with at least one of the number of target characters in the corresponding image, the area occupied by the target characters, the number of target logos, and the area occupied by the target logos.

5. The model training method according to any one of claims 1 to 3, wherein: The step of processing each first image to obtain at least one second image corresponding to each first image includes: Processing each first image to obtain a second image corresponding to each first image; The step of obtaining a corresponding filter weight parameter group according to each first image and at least one second image corresponding to each first image includes: According to each first image and a second image corresponding to each first image, a corresponding filter weight parameter group is obtained.

6. The model training method according to any one of claims 1 to 3, wherein: The processing of each first image includes at least one of adjusting the saturation of each first image, changing the brightness of each first image, blurring at least one area in each first image, and adding noise to each first image.

7. The model training method according to any one of claims 1 to 3, wherein: The filter includes a color lookup table.

8. An image processing method, wherein: The method comprises: Acquire at least one sampled image, and input the at least one sampled image into a trained filter weight parameter model to acquire a filter weight parameter group corresponding to each sampled image, wherein the filter weight parameter group includes filter weight parameters corresponding to at least two filters respectively, and the filter parameters of the at least two filters are different; wherein the trained filter weight parameter model is obtained by the method according to claim 1; the at least two filters include a first filter and a second filter, the first filter is used to adjust the pixel value of at least one pixel point in the corresponding image, and the second filter is used not to adjust the pixel value of any pixel point in the corresponding image; The filter parameters of the corresponding filter are adjusted according to the filter weight parameter group corresponding to each sampled image, and each test image is processed by the filter with the adjusted filter parameters to obtain at least one sampled processed image.

9. An electronic device comprising a memory and at least one processor; wherein: The memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the at least one processor to implement the method steps described in any one of claims 1-8.

10. A computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions, when executed by a processor, implement the method steps described in any one of claims 1 to 8.

Images