An image processing method, system, storage medium, and terminal device

Abstract

An embodiment of the present invention discloses an image processing method, system, storage medium, and terminal device, which are applied to the field of information processing technology based on artificial intelligence. The image processing system acquires the feature information of the image to be processed, maps the image to be processed to multiple types of degradation spaces to obtain degradation parameters based on each type of degradation control, then determines the adjustment information of the feature information based on the degradation parameters, and further adjusts the feature information according to the adjustment information to obtain the adjusted feature information. Finally, a high-resolution image of the image to be processed is obtained according to the adjusted feature information. During the process of processing the feature information of the image to be processed, it is guided by multiple types of degradation processes experienced by the image to be processed, so that the obtained adjusted feature information can pay more attention to multiple types of degradation, and the quality of the obtained high-resolution image is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information processing based on artificial intelligence, and particularly relates to an image processing method, system, storage medium and terminal device. Background Art

[0002] In the real world, low-resolution images are affected by various degradation processes (such as blur, noise, compression), and moreover, these degradations are mostly complex and unknown. When processing low-resolution images to obtain corresponding super-resolution images, it will exacerbate the complexity and difficulty of image processing.

[0003] Currently, when obtaining super-resolution images, it is mainly based on an independent degradation estimation network to process low-resolution images, so as to obtain corresponding high-resolution images, realizing a single mapping from low-resolution images to high-resolution images. In this process, only simple simulated degradation is considered, making the accuracy of degradation estimation greatly affect the quality of the obtained high-resolution images. Summary of the Invention

[0004] Embodiments of the present invention provide an image processing method, system, storage medium and terminal device, which improve the quality of the obtained high-resolution images.

[0005] On the one hand, an embodiment of the present invention provides an image processing method, including:

[0006] Obtaining feature information of an image to be processed;

[0007] Mapping the image to be processed to multiple types of degradation spaces to obtain degradation parameters of the image to be processed based on each type of degradation space;

[0008] Determining adjustment information of the feature information according to the degradation parameters of each type of degradation space;

[0009] Adjusting the feature information according to the adjustment information to obtain adjusted feature information;

[0010] Obtaining a high-resolution image of the image to be processed according to the adjusted feature information, and the resolution of the high-resolution image is higher than that of the image to be processed.

[0011] On the other hand, an embodiment of the present invention provides an image processing system, including:

[0012] A feature acquisition unit, configured to obtain feature information of an image to be processed;

[0013] A degradation parameter unit, configured to map the image to be processed to multiple types of degradation spaces to obtain degradation parameters of the image to be processed based on each type of degradation space;

[0014] An adjustment information unit, configured to determine adjustment information of the feature information according to degradation parameters of each type of degradation space;

[0015] An adjustment unit, configured to adjust the feature information according to the adjustment information to obtain adjusted feature information;

[0016] An image acquisition unit, configured to acquire a high-resolution image of the image to be processed according to the adjusted feature information, where the resolution of the high-resolution image is higher than that of the image to be processed.

[0017] Another aspect of the embodiments of the present invention further provides a computer-readable storage medium, which stores a plurality of computer programs, and the computer programs are adapted to be loaded and executed by a processor to perform the image processing method as described in one aspect of the embodiments of the present invention.

[0018] Another aspect of the embodiments of the present invention further provides a terminal device, including a processor and a memory;

[0019] The memory is used to store a plurality of computer programs, and the computer programs are used to be loaded and executed by a processor to perform the image processing method as described in one aspect of the embodiments of the present invention; the processor is used to implement each computer program in the plurality of computer programs.

[0020] It can be seen that in the method of this embodiment, the image processing system will acquire the feature information of the image to be processed, map the image to be processed to multiple types of degradation spaces to obtain degradation parameters based on each type of degradation control, then determine the adjustment information of the feature information based on the degradation parameters, and then adjust the feature information according to the adjustment information to obtain the adjusted feature information. Finally, a high-resolution image of the image to be processed is obtained according to the adjusted feature information. During the process of processing the feature information of the image to be processed, it is guided by multiple types of degradation processes experienced by the image to be processed, so that the obtained adjusted feature information can pay more attention to multiple types of degradation, and the quality of the obtained high-resolution image is improved to a large extent. Description of the Drawings

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a schematic diagram of an image processing method provided by an embodiment of the present invention;

[0023] Figure 2 It is a flowchart of an image processing method provided by an embodiment of the present invention;

[0024] Figure 3 It is a flowchart of a method for training an image processing model in another embodiment of the present invention;

[0025] Figure 4 It is a schematic diagram of an initial image processing model determined in another embodiment of the present invention;

[0026] Figure 5 It is a flowchart of a method for training an image processing model in an application embodiment of the present invention;

[0027] Figure 6 It is a schematic diagram of an initial image processing model determined in an application embodiment of the present invention;

[0028] Figure 7 It is a schematic diagram of a degradation prediction module in an application embodiment of the present invention;

[0029] Figure 8 It is a schematic diagram of a distributed system to which the image processing method is applied in another application embodiment of the present invention;

[0030] Figure 9 It is a schematic diagram of a block structure in another application embodiment of the present invention;

[0031] Figure 10 It is a schematic diagram of the logical structure of an image processing system provided by an embodiment of the present invention;

[0032] Figure 11 It is a schematic diagram of the logical structure of a terminal device provided by an embodiment of the present invention. Detailed implementation manners

[0033] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0034] In the description, claims and the above drawings of the present invention, the terms "first", "second", "third", "fourth", etc. (if any) are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0035] An embodiment of the present invention provides an image processing method, which mainly processes a low-resolution image to obtain a high-resolution image. Specifically, as Figure 1 shown, the image processing system can perform image processing by the following steps:

[0036] Obtain the feature information of the image to be processed;

[0037] Map the image to be processed to multiple types of degradation spaces to obtain the degradation parameters of the image to be processed based on each type of degradation space;

[0038] Determine the adjustment information of the feature information according to the degradation parameters of each type of degradation space;

[0039] Adjust the feature information according to the adjustment information to obtain the adjusted feature information;

[0040] Obtain the high-resolution image of the image to be processed according to the adjusted feature information, and the resolution of the high-resolution image is higher than that of the image to be processed.

[0041] In a specific application example, the image processing system can be applied to any application scenario, and specifically can be applied to, but not limited to, the following user terminals that need to process images: mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, etc.

[0042] Specifically, each step of the above image processing can be implemented by a pre-set image processing model, which is a machine learning model based on artificial intelligence. Among them, Artificial Intelligence (AI) uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results in terms of theory, methods, technologies, and application systems. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines to enable machines to have the functions of perception, reasoning, and decision-making.

[0043] Artificial intelligence technology is an interdisciplinary subject with a wide range of fields involved, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0044] Machine Learning (ML) is an interdisciplinary subject involving multiple fields such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent, and its applications cover all fields of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and learning from demonstration.

[0045] In this way, during the process of processing the feature information of the image to be processed, it is guided by multiple types of degradation processes that the image to be processed has experienced, so that the obtained adjusted feature information can pay more attention to multiple types of degradation, and to a large extent improve the quality of the obtained high-resolution image.

[0046] An embodiment of the present invention provides an image processing method, which is mainly a method executed by an image processing system. The flowchart is as Figure 2 shown and includes:

[0047] Step 101, obtain the feature information of the image to be processed.

[0048] It can be understood that in some application scenarios, when an image processing system needs to process a low-resolution image to obtain a corresponding high-resolution image, the low-resolution image can be used as the image to be processed, and the image processing process of this embodiment can be initiated.

[0049] Step 102: Map the image to be processed to multiple types of degradation spaces to obtain the degradation parameters of the image to be processed based on each type of degradation space.

[0050] Here, the degradation parameter of any type of degradation space is a parameter used to describe the corresponding type of degradation process experienced by the image to be processed. Specifically, it can be attributes involved in the degradation process such as the overall degradation score and the degree of degradation. Among them, the degradation process can include but is not limited to the following types: blur, noise, compression, etc. In this embodiment, a corresponding degradation parameter can be obtained for one type of degradation space.

[0051] Step 103: Determine the adjustment information of the feature information obtained in the above step 101 according to the degradation parameters of each type of degradation space.

[0052] It should be noted that there is no absolute order relationship between the above step 101 and steps 102 and 103. They can be executed simultaneously or sequentially. The one shown in the figure is just one specific application example.

[0053] Step 104: Adjust the feature information according to the adjustment information to obtain the adjusted feature information.

[0054] Specifically, the adjusted feature information can be obtained by performing a certain calculation on the adjustment information and the feature information, such as performing a calculation in the way of affine transformation.

[0055] In a specific embodiment, if the image processing system obtains the feature information of multiple levels of the image to be processed in the above step 101, when performing the above step 103, it is necessary to determine the adjustment information of each level according to the degradation parameters of each type of degradation space for each level. When performing this step 102, the feature information of the corresponding level will be adjusted according to the adjustment information of each level.

[0056] Among them, when obtaining the feature information of multiple levels of the image to be processed, after obtaining the feature information of one level, the image processing system will obtain the feature information of another level by combining the adjustment information of this level on the basis of the feature information of one level, and so on. After obtaining the feature information of the last level, the feature information of the last level can be adjusted by the adjustment information of the corresponding level to obtain the adjusted feature information, and based on this adjusted feature information, the following step 105 is executed. Among them, the feature information of one level can describe the information within a certain range of the image to be processed.

[0057] Step 105: Obtain a high-resolution image of the image to be processed according to the adjusted feature information. The resolution of the high-resolution image is higher than that of the image to be processed above.

[0058] In a specific embodiment, when the image processing system initiates the process of this embodiment, it can first call a preset image processing model. The image processing model includes a feature extraction module, a degradation prediction module, an adjustment generation module, and an output module. Therefore, when the image processing system executes the above step 102, it mainly maps the image to be processed to multiple types of degradation spaces through the degradation prediction module to obtain the degradation parameters of the image to be processed based on each type of degradation space; when executing the above step 103, the adjustment generation module determines the adjustment information of the feature information according to the degradation parameters of each type of degradation space; when executing the above steps 101 and 104, the feature extraction module is used to obtain the feature information of the image to be processed, and the feature information is adjusted according to the adjustment information to obtain the adjusted feature information; when executing the above step 105, the output module outputs the high-resolution image of the image to be processed according to the adjusted feature information.

[0059] It can be seen that in the method of this embodiment, the image processing system obtains the feature information of the image to be processed, maps the image to be processed to multiple types of degradation spaces to obtain the degradation parameters based on each type of degradation control, then determines the adjustment information of the feature information based on the degradation parameters, and then adjusts the feature information according to the adjustment information to obtain the adjusted feature information. Finally, a high-resolution image of the image to be processed is obtained according to the adjusted feature information. During the process of processing the feature information of the image to be processed, it is guided by multiple types of degradation processes experienced by the image to be processed, so that the obtained adjusted feature information can pay more attention to multiple types of degradation, and the quality of the obtained high-resolution image is improved to a large extent.

[0060] In a specific embodiment, the steps in the above steps 101 and 105 can be implemented by a preset image processing model, and the image processing model can be trained by the following method. The flowchart is as Figure 3 shown, including:

[0061] Step 201: Determine an initial image processing model. The initial image processing model includes a feature extraction module, a degradation prediction module, an adjustment generation module, and an output module.

[0062] It can be understood that when the image processing system determines the initial image processing model, it will determine the multi-layer structure included in the initial image processing model and the initial values of the parameters in each layer structure.

[0063] Specifically, as Figure 4As shown in the figure, the initial image processing model may include: a feature extraction module 10, a degradation prediction module 11, an adjustment generation module 12, and an output module 13, where: the feature extraction module 10 is used to extract the feature information of each low-resolution sample image; the degradation prediction module 11 is used to predict the degradation parameters of the low-resolution sample image based on multiple types of degradation spaces; the adjustment generation module 12 is used to determine the adjustment information of the feature information according to the degradation parameters obtained by the degradation prediction module 11, and input the adjustment information into the above-mentioned feature extraction module 10, then the feature extraction module 10 adjusts the feature information of the low-resolution sample image according to the adjustment information to obtain the adjusted feature information; the output module 13 outputs the high-resolution image of the low-resolution sample image according to the adjusted feature information.

[0064] Step 202: Determine the training samples, where the training samples include multiple low-resolution sample images and their corresponding high-resolution sample images.

[0065] Step 203: Obtain the high-resolution image corresponding to each low-resolution sample image through the initial image processing model.

[0066] Step 204: Adjust the initial image processing model according to the training samples and the high-resolution images obtained by the initial image processing model to obtain a preset image processing model.

[0067] Specifically, the image processing system will first calculate the first loss function and the second loss function related to the initial image processing model according to the results obtained by the initial image processing model in the above step 203 (i.e., the high-resolution images) and the high-resolution sample images in the training samples; then, according to the first loss function and the second loss function, adjust the parameter values in the feature extraction module 10, the adjustment generation module 12, and the output module 13 included in the initial image processing model. Specifically, an overall loss function can be calculated according to the first loss function and the second loss function, such as the weighted sum of the two loss functions, and then the parameter values in the feature extraction module 10, the adjustment generation module 12, and the output module 13 are adjusted based on the overall loss function. During the process of adjusting the parameter values based on the overall loss function, the parameter values in the above-mentioned degradation prediction module 11 can also be adjusted, which is not limited here. Among them:

[0068] The first loss function is used to indicate the difference between the high-resolution image obtained by the initial image processing model and the corresponding high-resolution sample image in the training sample (i.e., the actual high-resolution image of the low-resolution sample image), that is, the l1 loss function; the second loss function is used to indicate the difference between the feature information of each level obtained by the feature extraction module 10 and the feature information of the corresponding level in the corresponding high-resolution sample image in the training sample (i.e., the actual high-resolution image of the low-resolution sample image), that is, the perceptual loss function.

[0069] Further, a third loss function related to the initial image processing model can also be calculated. To calculate the third loss function, a discriminant module 14 needs to be included in the initial image processing model determined in the above step 201, which is connected to the above output module 13. The discriminant module 14 is used to determine whether the high-resolution sample image output by the output module 13 is real. In this way, the third loss function calculated by the initial image processing model includes the result of whether the high-resolution image output by the output module 13 obtained by the discriminant module 14 is real, and the expectation of the result of whether the discriminant module 14 determines that the actual high-resolution sample image (in the training samples) is real. The third loss function is an adversarial loss function; furthermore, when adjusting the parameter values in the feature extraction module 10, the adjustment generation module 12, and the output module 13 included in the initial image model, this third loss function can be combined, that is, the overall loss function is calculated based on the first loss function, the second loss function, and the third loss function, and then the parameter values are adjusted based on the calculated overall loss function.

[0070] It should be noted that during the training process of the feature extraction module 10, the adjustment generation module 12, and the output module 13 included in the image processing model, it is necessary to minimize the value of the above overall loss function as much as possible. This training process continuously optimizes the parameter values of the feature extraction module 10, the adjustment generation module 12, and the output module 13 included in the initial image processing model determined in the above step 201 through a series of mathematical optimization means such as backpropagation derivative and gradient descent, and makes the calculated value of the above overall loss function drop to the lowest. And during this process, the parameter values in the above degradation prediction module 11 can remain unchanged. Specifically, when the function value of the calculated overall loss function is relatively large, such as greater than a preset value, then the parameter values need to be changed, such as reducing the weight value of a neuron connection, etc., so that the function value of the overall loss function calculated according to the adjusted parameter values decreases.

[0071] Further, when training the parameter values in the degradation prediction module 11 included in the image processing model, during the training samples in the above step 202, each high-resolution sample image in the training samples needs to include a corresponding comparison sample group. The comparison sample group includes multiple low-resolution sample images obtained after the high-resolution sample image undergoes degradation processing. These multiple low-resolution sample images can be obtained based on different types of degradation spaces; then, during the process of the image processing system obtaining the high-resolution images corresponding to each low-resolution sample image through the initial image processing model in the above step 203, the degradation prediction module 11 can map each low-resolution sample image in the comparison sample group to multiple types of degradation spaces respectively, and obtain the degradation parameters of each low-resolution sample image based on each type of degradation space, that is, multiple degradation parameters can be obtained for one low-resolution sample image.

[0072] When the image processing system executes this step 204, specifically, it can calculate the first degradation loss function related to the degradation prediction module 11 according to the degradation parameters of each low-resolution sample image obtained by the degradation prediction module 11 of the initial image processing model. The first degradation loss function includes the distances between the degradation parameters of the low-resolution sample images based on each type of degradation space; then, it adjusts the parameter values in the degradation prediction module 11 according to the first degradation loss function.

[0073] For example, if the comparison sample group corresponding to a high-resolution sample image includes n low-resolution sample images, for each low-resolution sample image, m degradation parameters based on m types of degradation spaces can be obtained, that is, m degradation parameters, and each degradation parameter corresponds to a type of degradation space. In this way, when calculating the first degradation loss function, first determine that the number of low-resolution sample images belonging to each type of degradation space among the n low-resolution sample images is n1. Then, the degradation parameters of these n1 low-resolution sample images based on their respective types of degradation spaces are m1, m2,..., mn1. The first degradation loss function calculated includes the distances between any two degradation parameters in any type of degradation space, that is, the distances between any two of m1, m2,..., mn1.

[0074] It should be noted that the process of training the degradation prediction module 11 included in the image processing model is to minimize the value of the above first degradation loss function. This training process continuously optimizes the parameter values of the degradation prediction module 11 included in the initial image processing model determined in the above step 201 through a series of mathematical optimization means such as backpropagation derivation and gradient descent, and makes the calculated value of the above first degradation loss function drop to the lowest.

[0075] Furthermore, when training the parameter values in the degradation prediction module 11 included in the image processing model, in addition to using the above first degradation loss function to constrain the distances between the low-resolution sample images in each degradation space, the following second degradation loss function can also be used to constrain the range of the degradation parameters obtained by the degradation prediction module 11. Specifically:

[0076] When determining the training samples in the execution of step 202 above, it is necessary to include in the training samples the anchor point groups corresponding to each high-resolution sample image. The anchor point group includes the high-resolution sample image and a low-resolution sample image obtained after it has been subjected to multiple types of degradation processing. Moreover, when undergoing multiple types of degradation processing, the low-resolution sample image is obtained by degrading according to the maximum degradation intensity of each type of degradation space. Then, in the process of the image processing system executing step 203 above, it is necessary to map the high-resolution sample image and the low-resolution sample image in the anchor point group to multiple types of degradation spaces through the degradation prediction module 11 in the initial image processing model, so as to obtain the degradation parameters of the high-resolution sample image and the low-resolution sample image respectively based on each type of degradation space. That is, multiple degradation parameters are obtained for the low-resolution sample image, and multiple degradation parameters are also obtained for the high-resolution sample image.

[0077] Then, when the image processing system executes this step 204, it can specifically calculate the second degradation loss function related to the degradation prediction module 11 according to the degradation parameters corresponding to each low-resolution sample image and high-resolution sample image obtained by the degradation prediction module 11 of the initial image processing model. The second degradation loss function includes the calculated values of the degradation parameters of the high-resolution sample images in each anchor point group, and the calculated values of the degradation parameters of the low-resolution sample images in each anchor point group. Then, the parameter values in the degradation prediction module 11 are adjusted according to the second degradation loss function.

[0078] For example, if the anchor point group corresponding to a high-resolution sample image includes the high-resolution sample image and a low-resolution sample image, for each low-resolution sample image, degradation parameters a1, a2, ……, am based on m types of degradation spaces can be obtained, and for each high-resolution sample image, degradation parameters b1, b2, ……, bm based on m types of degradation spaces can be obtained. Each degradation parameter corresponds to a type of degradation space. In this way, when calculating the second degradation loss function, the second degradation loss function includes the calculated values of the degradation parameters a1, a2, ……, am of the low-resolution sample images, and the calculated values of the degradation parameters b1, b2, ……, bm of the high-resolution sample images.

[0079] It should be noted that in the process of training the degradation prediction module 11 included in the image processing model, it is also necessary to minimize the value of the above-mentioned second degradation loss function. This training process continuously optimizes the parameter values of the parameters in the degradation prediction module 11 included in the initial image processing model determined in step 201 above through a series of mathematical optimization means such as backpropagation derivative and gradient descent, and makes the calculated value of the above-mentioned second degradation loss function drop to the lowest.

[0080] In addition, it should be noted that the above steps 203 to 204 are an adjustment of the parameter values in the initial image processing model for each high-resolution image corresponding to the low-resolution sample images obtained by the initial image processing model. In practical applications, the above steps 203 to 204 need to be continuously looped until the adjustment of the parameter values meets a certain stop condition.

[0081] Therefore, after the image processing system executes the above steps 201 to 204 of the embodiment, it is also necessary to determine whether the current adjustment of the parameter values meets the preset stop condition. When it is met, the process ends, and the parameter values adjusted in the above step 204 are used as the parameter values in the finally trained image processing model. When the image processing model is preset in the image processing system, it may only include the trained feature extraction module 10, degradation prediction module 11, adjustment generation module 12, and output module 13, and does not include the above discriminant module 14; when it is not met, for the initial image processing model after adjusting the parameter values, return to execute the above steps 203 to 204. Among them, the preset stop conditions include but are not limited to any one of the following conditions: the difference between the currently adjusted parameter values and the parameter values adjusted last time is less than a threshold, that is, the adjusted parameter values reach convergence; and the number of times of adjusting the parameter values is equal to the preset number of times, etc.

[0082] The following uses a specific application example to illustrate the image processing method of the present invention. The method of this embodiment mainly includes the following two parts:

[0083] (1) Specifically, as Figure 5 shown, the image processing system can train the image processing model through the following steps:

[0084] Step 301, determine the structure of each layer included in the initial image processing model and the initial values of the parameters in each layer structure.

[0085] Specifically, as Figure 6 shown, the initial image processing model determined by the image processing system includes a feature extraction module 20, a degradation prediction module (Unsupervised Degradation Estimation Modul, UDEM) 21, an adjustment generation module 22, an output module 23, and a discriminant module 24, where:

[0086] The adjustment generation module 22 may include multiple cascaded levels. Each level can obtain a set of adjustment information for adjusting the feature information of one level. Each level may include a fully connected layer (FC) and a channelwise split. Specifically, the adjustment information of each level may include α and β for performing affine transformation.

[0087] The feature extraction module 20 includes multiple convolutional layers and multiple cascaded levels. Each level can obtain the feature information of that level and adjust the feature information. Specifically, each level can include a residual dense block and an adjustment sub-module to achieve this. The feature information of that level can be obtained through the residual dense block, and then the multiplier and adder in the adjustment sub-module can combine the adjustment information of the corresponding level obtained by the above adjustment generation module 22 to perform an affine transformation on the feature information of that level.

[0088] Step 302: Determine the training samples. The training samples can include a comparison sample group and an anchor group corresponding to multiple high-resolution sample images respectively. In the comparison sample group, it includes multiple low-resolution sample images obtained by degrading the high-resolution sample images. In the anchor group, it includes the high-resolution sample image and a low-resolution sample image obtained by performing various types of degradation processing according to the maximum degradation intensity of each type of degradation space.

[0089] In this embodiment, taking the degradation processing of two types, namely blur and noise, as an example for illustration, as Figure 7 shown, when generating the comparison sample group gc of each high-resolution sample image, a high-definition image I can be first subjected to complex and high-order degradation erosion to simulate the low-resolution images obtained by degradation in the real world, and a comparison sample group including three low-resolution sample images c1, c2, and c3 is obtained. The blur intensity of the low-resolution sample image c1 is greater than that of the low-resolution sample image c2, and the noise intensity of the low-resolution sample image c3 is greater than that of the low-resolution sample image c2. In the anchor group gd, the high-resolution sample image d2 has no noise degradation and blur degradation, while the low-resolution sample image d1 has undergone the maximum-intensity noise degradation and maximum-intensity blur degradation.

[0090] Step 303: Respectively obtain the high-resolution images corresponding to each low-resolution sample image through the initial image processing model.

[0091] Specifically, the degradation prediction module 21 predicts the degradation parameters of each low-resolution sample image based on multiple types of degradation spaces (in this embodiment, noise degradation and blur degradation are taken as examples for illustration); the adjustment generation module 22 determines the adjustment information corresponding to the feature information of multiple levels according to the degradation parameters obtained by the degradation prediction module 21, specifically α and β, and inputs the adjustment information of each level into the adjustment sub-module of each level in the above feature extraction module 10; the feature extraction module 20 first extracts the feature information of multiple levels of each low-resolution sample image, and then each level of adjustment sub-module adjusts the feature information of this level through the adjustment information of this level, and the adjusted feature information of the low-resolution sample is output by the adjustment sub-module of the last level; the output module 23 outputs the high-resolution image of the low-resolution sample image according to the adjusted feature information, and the discrimination module 24 determines whether the high-resolution image output by the output module 23 is real.

[0092] In this process, the degradation prediction module 21 can also predict the degradation parameters of the high-resolution sample images in each anchor point group based on multiple types of degradation spaces (in this embodiment, noise degradation and blur degradation are taken as examples for illustration).

[0093] Step 304, calculate the first loss function, the second loss function, and the third loss function, and then calculate the overall loss function according to these three loss functions, and adjust the parameter values of the feature extraction module 20, the adjustment generation module 22, and the output module 23 in the initial image processing model according to the overall loss function.

[0094] Specifically, the first loss function L 1 is used to indicate the difference between the high-resolution image G(x i ) output by the above output module 23 and the actual high-resolution sample image y in the training sample, and can be specifically represented by the following formula 1:

[0095]

[0096] The second loss function L 2 is used to indicate the difference between the feature information φ LR of each level i of each (denoted by j for the j-th) low-resolution sample image I i (G(I LR )) x,y obtained by the feature extraction module and the feature information φ HR of the corresponding high-resolution sample image I i,j in the training sample at the corresponding level i HR (I x,y ), and can be specifically represented by the following formula 1:

[0097]

[0098] The third loss function L includes the expectation of the result obtained by the discrimination module 24 on whether the high-resolution image output by the output module 23 is true. And the expectation of the result obtained by the discrimination module 24 on whether the actual high-resolution sample image is true. Specifically, it can be represented by the following formula 3:

[0099]

[0100] Step 305, calculate the first degradation loss function, and adjust the parameter values of the degradation prediction module 21 in the initial image processing model according to the first loss function.

[0101] Specifically, as Figure 7 shown, for the degradation parameters corresponding to each low-resolution sample image in each comparison sample group and anchor point group obtained by the degradation prediction module 21, and the degradation parameters corresponding to the high-resolution sample images in each anchor point group, it can be specifically represented by the following formula 4:

[0102]

[0103] The calculated first degradation loss function includes a loss function obtained based on the distance between any two degradation parameters in the noise degradation space and another loss function obtained based on the distance between any two degradation parameters in the blur degradation space. The distance between the low-resolution sample images in each degradation space is constrained by the first degradation loss function. Specifically, it can be represented by the following formula 5:

[0104]

[0105] Step 306, calculate the second degradation loss function, and adjust the parameter values of the degradation prediction module 21 in the initial image processing model according to the second loss function.

[0106] The second degradation loss function L AC includes a loss function obtained based on the calculated values of the degradation parameters of the low-resolution sample images in the anchor point group and another loss function obtained based on the calculated values of the degradation parameters of the high-resolution sample images in the anchor point group. The range of the degradation parameters obtained by the degradation prediction module is constrained by the second degradation loss function, so that the minimum value of the obtained degradation parameters is 0 and the maximum value is 1. Specifically, it can be represented by the following formula 6:

[0107]

[0108] Step 307: Determine whether the adjustment of the parameter values of each module meets a preset stop condition. If it meets, end the process, use the adjusted parameter values in the above steps as the parameter values in the trained image processing model, and preset the trained image processing model into the image processing system. The image processing model may include the above-mentioned feature extraction module 20, degradation prediction module 21, adjustment generation module 22, and output module 23. If it does not meet, return to execute step 303 above.

[0109] (2) When it is determined that the initial image processing model needs to process a low-resolution image to be processed, the preset image processing model in the system can be called first. This image processing model can directly process the image to be processed and output the corresponding high-resolution image.

[0110] In the practical process, after training the image processing model using the existing network structures (such as ESRGAN, CUGAN, and BSRGAN) and the network structure in the embodiment of the present invention on two datasets (i.e., Real SRSet and AIM19), multiple evaluation parameters are calculated, such as the no-reference image quality (NIQE), learned perceptual image patch similarity (LPIPS), and DISTS, as shown in Table 1 below:

[0111]

[0112] Table 1

[0113] The optimal parameter values of each evaluation parameter are shown in black font in Table 1. That is, the smaller the parameter value of each evaluation parameter, the better the quality of the trained image processing model. It can be seen that the quality of the high-resolution image obtained after processing the low-resolution image by the image processing model trained in the way of the embodiment of the present invention on the dataset AIM19 can be well improved.

[0114] The following uses another specific application example to illustrate the image processing method in the present invention. The image processing system in the embodiment of the present invention is mainly a distributed system 100. The distributed system may include a client 300 and multiple nodes 200 (any form of computing device connected to the network, such as a server, user terminal), and the client 300 is connected to the nodes 200 in the form of network communication.

[0115] Taking the distributed system as a blockchain system as an example, see Figure 8FIG. 0 is an optional structural diagram of the distributed system 100 provided by an embodiment of the present invention applied to a blockchain system, formed by multiple nodes 200 (any form of computing device accessing the network, such as a server, a user terminal) and a client 300. A peer-to-peer (P2P) network is formed among the nodes. The P2P protocol is an application layer protocol running on top of the Transmission Control Protocol (TCP). In a distributed system, any machine such as a server or a terminal can join and become a node. A node includes a hardware layer, an intermediate layer, an operating system layer, and an application layer.

[0116] See Figure 8 the functions of each node in the blockchain system shown, and the functions involved include:

[0117] 1) Routing, a basic function of a node, used to support communication between nodes.

[0118] In addition to the routing function, a node may also have the following functions:

[0119] 2) Application, which is used to be deployed in the blockchain, implement specific services according to actual business requirements, record data related to the implemented functions to form record data, carry a digital signature in the record data to indicate the source of the task data, and send the record data to other nodes in the blockchain system. When other nodes verify the source and integrity of the record data successfully, the record data is added to a temporary block.

[0120] For example, the service implemented by the application includes code for implementing an image processing function, and the image processing function mainly includes:

[0121] Obtaining feature information of an image to be processed; mapping the image to be processed to multiple types of degradation spaces to obtain degradation parameters of the image to be processed based on each type of degradation space; determining adjustment information of the feature information according to the degradation parameters of each type of degradation space; adjusting the feature information according to the adjustment information to obtain adjusted feature information; and obtaining a high-resolution image of the image to be processed according to the adjusted feature information, where the resolution of the high-resolution image is higher than that of the image to be processed.

[0122] 3) Blockchain, including a series of blocks (Blocks) that are sequentially connected in the order of generation. Once a new block is added to the blockchain, it will not be removed again. The blocks record the record data submitted by the nodes in the blockchain system.

[0123] See Figure 9An optional schematic diagram of the block structure provided by the embodiments of the present invention. Each block includes the hash value of the transaction records stored in this block (the hash value of this block) and the hash value of the previous block. Each block is connected through the hash value to form a blockchain. In addition, the block may further include information such as the timestamp when the block is generated. A blockchain, in essence, is a decentralized database, a series of data blocks generated by using cryptographic methods. Each data block contains relevant information for verifying the validity of its information (anti-counterfeiting) and generating the next block.

[0124] The embodiments of the present invention also provide an image processing system, and its schematic structural diagram is as Figure 10 shown, and specifically may include:

[0125] A feature acquisition unit 30, configured to acquire the feature information of the image to be processed;

[0126] A degradation parameter unit 31, configured to map the image to be processed to multiple types of degradation spaces to obtain the degradation parameters of the image to be processed based on each type of degradation space.

[0127] An adjustment information unit 32, configured to determine the adjustment information of the feature information according to the degradation parameters of each type of degradation space obtained by the degradation parameter unit 31.

[0128] An adjustment unit 33, configured to adjust the feature information acquired by the feature acquisition unit 30 according to the adjustment information determined by the adjustment information unit 32 to obtain the adjusted feature information.

[0129] An image acquisition unit 34, configured to acquire the high-resolution image of the image to be processed according to the adjusted feature information obtained by the adjustment unit 33, and the resolution of the high-resolution image is higher than the resolution of the image to be processed.

[0130] In a specific embodiment, the above-mentioned feature acquisition unit 30 is specifically configured to acquire the feature information of multiple levels of the image to be processed; the degradation parameter unit 31 is specifically configured to determine the adjustment information of each level according to the degradation parameters of each type of degradation space; the adjustment information unit 32 is specifically configured to adjust the feature information of the corresponding level according to the adjustment information of each level.

[0131] In another specific embodiment, the image processing system further includes: a calling unit 35, configured to call a pre-set image processing model, where the image processing model includes a feature extraction module, a degradation prediction module, an adjustment generation module, and an output module; thus, in the degradation parameter unit 31, the degradation prediction module maps the image to be processed into multiple types of degradation spaces, and obtains the degradation parameters of the image to be processed based on each type of degradation space; in the adjustment information unit 32, the adjustment generation module determines the adjustment information of the feature information according to the degradation parameters of each type of degradation space; in the feature acquisition unit 30, the feature extraction module acquires the feature information of the image to be processed, and adjusts the feature information according to the adjustment information to obtain adjusted feature information; in the image acquisition unit 34, the output module outputs a high-resolution image of the image to be processed according to the adjusted feature information.

[0132] Further, the image processing system may further include:

[0133] a training unit 36, configured to determine an initial image processing model, where the initial image processing model includes a feature extraction module, a degradation prediction module, an adjustment generation module, and an output module; determine training samples, where the training samples include a plurality of low-resolution sample images and their corresponding high-resolution sample images respectively; acquire the high-resolution images corresponding to each low-resolution sample image through the initial image processing model; and adjust the initial image processing model according to the training samples and the high-resolution images obtained by the initial image processing model, so as to obtain the pre-set image processing model.

[0134] Wherein, when the training unit 36 adjusts the initial image processing model according to the training samples and the high-resolution images obtained by the initial image processing model, it is specifically configured to calculate a first loss function and a second loss function related to the initial image processing model; wherein, the first loss function is used to indicate the difference between the high-resolution image obtained by the initial image processing model and the corresponding high-resolution sample image in the training samples, and the second loss function is used to indicate the difference between the feature information of each level acquired by the feature extraction module and the feature information of the corresponding high-resolution sample image in the training samples at the corresponding level; and adjust the parameter values in the feature extraction module, the adjustment generation module, and the output module included in the initial image model according to the first loss function and the second loss function.

[0135] Among them, if the initial image processing model further includes a discrimination module connected to the output module, the training unit 36 is further configured to calculate a third loss function related to the initial image processing model when adjusting the initial image processing model. The third loss function includes: the result of whether the high-resolution image output by the output module obtained by the discrimination module is true, and the expectation of the result of the discrimination module determining whether the high-resolution sample image in the training sample is true. Therefore, according to the first loss function and the second loss function, adjusting the parameter values in the feature extraction module, the adjustment generation module, and the output module included in the initial image model specifically includes: adjusting the parameter values in the feature extraction module, the adjustment generation module, and the output module included in the initial image model according to the first loss function, the second loss function, and the third loss function.

[0136] Further, if the training sample includes a comparison sample group corresponding to each high-resolution sample image, and the comparison sample group includes multiple low-resolution sample images obtained by degrading the high-resolution sample image; when obtaining the high-resolution image corresponding to each low-resolution sample image through the initial image processing model, the degradation prediction module maps each low-resolution sample image in the comparison sample group to multiple types of degradation spaces respectively to obtain the degradation parameters of each low-resolution sample image based on each type of degradation space. In this way, when the training unit 36 adjusts the initial image processing model according to the high-resolution image obtained from the training sample and the initial image processing model, it is specifically configured to calculate a first degradation loss function related to the degradation prediction module according to the degradation parameters of each low-resolution sample image obtained by the degradation prediction module of the initial image processing model. The first degradation loss function includes the distances between the degradation parameters of the low-resolution sample images based on each type of degradation space. Adjust the parameter values in the degradation prediction module according to the first degradation loss function.

[0137] Further, if the training samples further include: an anchor group corresponding to each high-resolution sample image, where the anchor group includes a high-resolution sample image and a low-resolution sample image obtained after it undergoes multiple types of degradation processing; the degradation prediction module in the initial image processing model maps the high-resolution sample image and the low-resolution sample image in the anchor group to multiple types of degradation spaces respectively, to obtain the degradation parameters of the high-resolution sample image and the low-resolution sample image respectively based on each type of degradation space; in this way, when the training unit 36 adjusts the initial image processing model according to the high-resolution image obtained from the training samples and the initial image processing model, it is specifically configured to calculate a second degradation loss function related to the degradation prediction module according to the degradation parameters corresponding to each low-resolution sample image and high-resolution sample image obtained by the degradation prediction module of the initial image processing model; the second degradation loss function includes the calculated values of the degradation parameters of the high-resolution sample images in each anchor group and the calculated values of the degradation parameters of the low-resolution sample images in each anchor group; and adjust the parameter values in the degradation prediction module according to the second degradation loss function.

[0138] Further, the training unit 36 is further configured to stop adjusting the parameter values when the number of times of adjusting the parameter values is equal to a preset number of times, or if the difference between the currently adjusted parameter value and the parameter value adjusted last time is less than a threshold.

[0139] It can be seen that in the image processing system of this embodiment, the feature acquisition unit 30 acquires the feature information of the image to be processed, and the degradation parameter unit 31 maps the image to be processed to multiple types of degradation spaces to obtain the degradation parameters based on each type of degradation control, and then the adjustment information unit 32 determines the adjustment information of the feature information based on the degradation parameters, and further the adjustment unit 33 adjusts the feature information according to the adjustment information to obtain the adjusted feature information, and finally the image acquisition unit 34 obtains the high-resolution image of the image to be processed according to the adjusted feature information. During the process of processing the feature information of the image to be processed, it is guided by multiple types of degradation processes experienced by the image to be processed, so that the obtained adjusted feature information can pay more attention to multiple types of degradation, and the quality of the obtained high-resolution image is improved to a large extent.

[0140] The embodiment of the present invention further provides a terminal device, and its structural schematic diagram is as Figure 11As shown, the terminal device can vary significantly due to different configurations or performances, and may include one or more central processing units (CPUs) 40 (e.g., one or more processors) and a memory 41, and one or more storage media 42 (e.g., one or more mass storage devices) for storing application programs 421 or data 422. Among them, the memory 41 and the storage media 42 can be transient storage or persistent storage. The programs stored in the storage media 42 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations for the terminal device. Further, the central processing unit 40 can be configured to communicate with the storage media 42 and execute a series of instruction operations in the storage media 42 on the terminal device.

[0141] Specifically, the application program 421 stored in the storage media 42 includes an application program for image processing, and this program may include the feature acquisition unit 30, the degradation parameter unit 31, the adjustment information unit 32, the adjustment unit 33, the image acquisition unit 34, the call unit 35, and the training unit 36 in the above-mentioned image processing system, which will not be elaborated here. Further, the central processing unit 40 can be configured to communicate with the storage media 42 and execute a series of operations corresponding to the image processing application program stored in the storage media 42 on the terminal device.

[0142] The terminal device may further include one or more power supplies 43, one or more wired or wireless network interfaces 44, one or more input / output interfaces 45, and / or one or more operating systems 423, such as WindowsServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0143] The steps performed by the image processing system in the above method embodiments may be based on the Figure 11 structure of the shown terminal device.

[0144] Further, on the other hand, an embodiment of the present invention also provides a computer-readable storage medium, which stores a plurality of computer programs, and the computer programs are adapted to be loaded and executed by a processor to perform the image processing method performed by the above-mentioned image processing system.

[0145] On the other hand, an embodiment of the present invention also provides a terminal device, including a processor and a memory;

[0146] The memory is used to store a plurality of computer programs, and the computer programs are used to be loaded and executed by a processor to perform an image processing method as executed by the above-mentioned image processing system; the processor is used to implement each of the plurality of computer programs.

[0147] In addition, according to an aspect of the present application, there is provided a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the image processing method provided in the above various optional implementation manners.

[0148] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. The storage medium may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disc, etc.

[0149] The above has introduced in detail an image processing method, system, storage medium and terminal device provided by the embodiments of the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An image processing method, characterized in that, it includes: Obtaining the feature information of the image to be processed; Mapping the image to be processed into multiple types of degradation spaces to obtain the degradation parameters of the image to be processed based on each type of degradation space; wherein, the degradation parameter is a parameter used to describe the corresponding type of degradation process experienced by the image to be processed; Determining the adjustment information of the feature information according to the degradation parameters of each type of degradation space; Adjusting the feature information according to the adjustment information to obtain the adjusted feature information; Obtaining the high-resolution image of the image to be processed according to the adjusted feature information, and the resolution of the high-resolution image is higher than that of the image to be processed.

2. The method according to claim 1, characterized in that, The obtaining of the feature information of the image to be processed specifically includes: obtaining the feature information of multiple levels of the image to be processed; The determining of the adjustment information of the feature information according to the degradation parameters of each type of degradation space specifically includes: determining the adjustment information of each level according to the degradation parameters of each type of degradation space; The adjusting of the feature information according to the adjustment information to obtain the adjusted feature information specifically includes: adjusting the feature information of the corresponding level according to the adjustment information of each level.

3. The method according to claim 1, characterized in that, Before obtaining the feature information of the image to be processed, it further includes: calling a preset image processing model, and the image processing model includes a feature extraction module, a degradation prediction module, an adjustment generation module, and an output module; The mapping of the image to be processed into multiple types of degradation spaces to obtain the degradation parameters of the image to be processed based on each type of degradation space specifically includes: mapping the image to be processed into multiple types of degradation spaces through the degradation prediction module to obtain the degradation parameters of the image to be processed based on each type of degradation space; The determining of the adjustment information of the feature information according to the degradation parameters of each type of degradation space specifically includes: the adjustment generation module determines the adjustment information of the feature information according to the degradation parameters of each type of degradation space; The obtaining of the feature information of the image to be processed, adjusting the feature information according to the adjustment information to obtain the adjusted feature information specifically includes: obtaining the feature information of the image to be processed through the feature extraction module, and adjusting the feature information according to the adjustment information to obtain the adjusted feature information; The obtaining of the high-resolution image of the image to be processed according to the adjusted feature information specifically includes: the output module outputs the high-resolution image of the image to be processed according to the adjusted feature information.

4. The method according to claim 3, characterized in that, Determine the initial image processing model, and the initial image processing model includes a feature extraction module, a degradation prediction module, an adjustment generation module, and an output module; Determine the training samples, and the training samples include multiple low-resolution sample images and their respectively corresponding high-resolution sample images; Obtain the high-resolution images corresponding to each low-resolution sample image through the initial image processing model respectively; Adjust the initial image processing model according to the training samples and the high-resolution images obtained by the initial image processing model, so as to obtain the preset image processing model.

5. The method according to claim 4, characterized in that, The adjusting the initial image processing model according to the training samples and the high-resolution images obtained by the initial image processing model specifically includes: Calculating a first loss function and a second loss function related to the initial image processing model; wherein, the first loss function is used to indicate the difference between the high-resolution image obtained by the initial image processing model and the corresponding high-resolution sample image in the training samples, and the second loss function is used to indicate the difference between the feature information of each level obtained by the feature extraction module and the feature information of the corresponding high-resolution sample image in the training samples at the corresponding level; Adjust the parameter values in the feature extraction module, adjustment generation module and output module included in the initial image model according to the first loss function and the second loss function.

6. The method according to claim 5, characterized in that, The initial image processing model further includes a discriminant module, which is connected to the output module; Then the adjusting the initial image processing model according to the training samples and the high-resolution images obtained by the initial image processing model further includes: Calculating a third loss function related to the initial image processing model, where the third loss function includes: the result of whether the high-resolution image output by the output module obtained by the discriminant module is true, and the expectation of the result of the discriminant module discriminating whether the high-resolution sample image in the training samples is true; Therefore, adjusting the parameter values in the feature extraction module, adjustment generation module and output module included in the initial image model according to the first loss function and the second loss function specifically includes: adjusting the parameter values in the feature extraction module, adjustment generation module and output module included in the initial image model according to the first loss function, the second loss function and the third loss function.

7. The method according to claim 4, characterized in that, The training samples include: a comparison sample group corresponding to each high-resolution sample image respectively, and the comparison sample group includes a plurality of low-resolution sample images obtained after the high-resolution sample image is degraded; When obtaining the high-resolution images corresponding to each low-resolution sample image through the initial image processing model, the degradation prediction module maps each low-resolution sample image in the comparison sample group to a plurality of types of degradation spaces respectively, so as to obtain the degradation parameters of each low-resolution sample image based on each type of degradation space; The adjusting the initial image processing model according to the training samples and the high-resolution images obtained by the initial image processing model specifically includes: Calculate a first degradation loss function related to the degradation prediction module according to the degradation parameters of each low-resolution sample image obtained by the degradation prediction module of the initial image processing model; the first degradation loss function includes the distances between the degradation parameters of the low-resolution sample images in each type of degradation space. Adjust the parameter values in the degradation prediction module according to the first degradation loss function.

8. The method according to claim 7, wherein, the training samples further include: an anchor point group corresponding to each high-resolution sample image, and the anchor point group includes a high-resolution sample image and a low-resolution sample image obtained after the high-resolution sample image is subjected to multiple types of degradation processing; mapping the high-resolution sample image and the low-resolution sample image in the anchor point group to multiple types of degradation spaces respectively through the degradation prediction module in the initial image processing model, to obtain the degradation parameters of the high-resolution sample image and the low-resolution sample image respectively based on each type of degradation space; adjusting the initial image processing model according to the high-resolution image obtained from the training samples and the initial image processing model, specifically including: Calculate a second degradation loss function related to the degradation prediction module according to the degradation parameters corresponding to each low-resolution sample image and high-resolution sample image obtained by the degradation prediction module of the initial image processing model; the second degradation loss function includes the calculated values of the degradation parameters of the high-resolution sample images in each anchor point group, and the calculated values of the degradation parameters of the low-resolution sample images in each anchor point group; Adjust the parameter values in the degradation prediction module according to the second degradation loss function.

9. The method according to any one of claims 5 to 8, wherein, When the number of times of adjusting the parameter values is equal to a preset number of times, or if the difference between the currently adjusted parameter value and the parameter value adjusted last time is less than a threshold value, then stop adjusting the parameter values.

10. An image processing system, wherein, including: a feature acquisition unit, configured to acquire feature information of an image to be processed; a degradation parameter unit, configured to map the image to be processed to multiple types of degradation spaces, to obtain the degradation parameters of the image to be processed based on each type of degradation space; wherein, the degradation parameter is a parameter used to describe the corresponding type of degradation process experienced by the image to be processed; an adjustment information unit, configured to determine adjustment information of the feature information according to the degradation parameters of each type of degradation space; an adjustment unit, configured to adjust the feature information according to the adjustment information to obtain adjusted feature information; an image acquisition unit, configured to acquire a high-resolution image of the image to be processed according to the adjusted feature information, and the resolution of the high-resolution image is higher than the resolution of the image to be processed.

11. A computer-readable storage medium, wherein, The computer-readable storage medium stores multiple computer programs, and the computer programs are adapted to be loaded and executed by a processor to perform the image processing method according to any one of claims 1 to 9.

12. A terminal device, wherein, including a processor and a memory; the memory is used for storing a plurality of computer programs, and the computer programs are used to be loaded and executed by the processor to perform the image processing method according to any one of claims 1 to 9; the processor is used to implement each of the plurality of computer programs.

13. A computer program product, characterized in that, the computer program product includes computer instructions, and the computer instructions are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the image processing method according to any one of claims 1 to 9.

Images