Image Denoising Method, Device, Computer Storage Medium and Electronic Device

Abstract

The present application discloses an image denoising method, apparatus, computer storage medium and electronic device. The method includes: obtaining a composite convex minimization model for performing image denoising on a target image, where the composite convex minimization model includes an objective function with the minimum noise of the target image as the objective and an equality constraint condition for variables in the objective function; determining an update rule for variables required to solve the composite convex minimization model, and determining a closed-form solution for the pixel values of pixel points in the target image, where the closed-form solution includes second-order derivative information of the objective function; iteratively solving the pixel values of each pixel point in the target image according to the closed-form solution to obtain the target pixel values of each pixel point in the case of the minimum noise of the target image, and forming a denoised image. Through the present application, the problem of large computational complexity in accurately solving the pixel values of the denoised image when image denoising is a composite convex minimization problem in the related art is solved.

Description

Technical Field

[0001] The present application relates to the field of image denoising, and more particularly, to an image denoising method, apparatus, computer storage medium, and electronic device. Background Art

[0002] In the field of engineering technology, such as the field of image processing, if the image denoising problem is abstracted as the following composite convex minimization problem: where f(x) is a loss function, Ax + By = c is the equality constraint condition of this minimization problem, where A and B are given matrices, and A is a row full-rank matrix, c is a constant vector, and g(y) is a regularization term function. x refers to the image, n is the number of pixel points in x, and f i (x) is the loss function at pixel point i, and they are all convex functions, continuously differentiable, and have L i -Lipschitz continuous gradients. It is necessary to solve the values of x and y when f(x) + g(y) is minimized under the constraints to determine the pixel values of the denoised image.

[0003] To solve the above problem, the related technology has emerged with the variance-reduced stochastic ADMM method. This method has the advantages of being easy to implement and convenient for parallelization. However, as a first-order gradient method, it has obvious disadvantages, that is, its performance is very sensitive to the conditions of the target problem. Limited by ill-conditioned conditions, its robustness (i.e., anti-disturbance ability) is also not good. Specifically, when the condition number (L and λ are the Lipschitz continuous gradient coefficient (also called the smooth coefficient) and the strong convex coefficient of the loss function f, respectively) of the image denoising problem is very large, under specific conditions, the actual performance of this scheme will deteriorate significantly. It may take thousands of iterations to reach a four-digit accuracy to accurately solve the pixel values of the denoised image.

[0004] Aiming at the problem of large computational complexity in accurately solving the pixel values of the denoised image when the image denoising is a composite convex minimization problem in the related technology, no effective solution has been proposed yet. Summary of the Invention

[0005] The present application provides an image denoising method, apparatus, computer storage medium, and electronic device to solve the problem of large computational complexity in accurately solving the pixel values of the denoised image when the image denoising is a composite convex minimization problem in the related technology.

[0006] According to one aspect of the present application, an image denoising method is provided. The method includes: obtaining a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function aiming at minimizing the noise of the target image and equality constraint conditions of variables in the objective function; determining an update rule for variables required to solve the composite convex minimization model, and determining a closed-form solution for the pixel values of pixel points in the target image based on the required variable update rule, where the closed-form solution includes second-order derivative information of the objective function; iteratively solving the pixel values of each pixel point in the target image according to the closed-form solution to obtain the target pixel values of each pixel point in the case of minimizing the noise of the target image, and forming a denoised image from the obtained target pixel values of each pixel point.

[0007] Optionally, the objective function includes a loss function and a regularization term function. The variable of the loss function is the pixel value of the pixel point in the target image, and the variable of the regularization term function is a regularization term variable. Determining the update rule for variables required to solve the composite convex minimization model includes: determining an augmented Lagrangian function according to the composite convex minimization model, where the variables of the augmented Lagrangian function include: the pixel value of the pixel point in the target image, the regularization term variable, and the dual variable; minimizing the augmented Lagrangian function with respect to the pixel value of the pixel point to obtain an update rule for the pixel value of the pixel point in the target image; minimizing the augmented Lagrangian function with respect to the regularization term variable to obtain an update rule for the regularization term variable; minimizing the augmented Lagrangian function with respect to the dual variable to obtain an update rule for the dual variable.

[0008] Optionally, the closed-form solution for the pixel value of the pixel point in the target image is used to characterize the relationship between the pixel value of the current pixel point and the pixel value obtained in the previous iteration. Iteratively solving the pixel values of each pixel point in the target image according to the closed-form solution to obtain the target pixel values of each pixel point in the case of minimizing the noise of the target image includes: alternately iteratively solving the pixel value, the regularization term variable, and the dual variable according to the closed-form solution for the pixel value of the pixel point in the target image, the update rule for the regularization term variable, and the update rule for the dual variable until the iteration preset number is reached to obtain the target pixel value of each pixel point.

[0009] Optionally, the objective function includes a loss function and a regularization term function. The regularization term function is determined by the norm of the product of the pixel value of the pixel point in the target image and a positive definite matrix, and the loss function is determined by the norm of the noise of the pixel point in the target image. The noise of the pixel point in the target image is determined by the difference between the pixel value of the pixel point and the pixel value of the pixel point in the original image.

[0010] Optionally, the closed-form solution includes a preprocessor, and the preprocessor includes second-order derivative information of the objective function.

[0011] Optionally, the preprocessor is obtained by taking the second derivative of the objective function to obtain the second derivative result, and multiplying the second derivative result by a preset coefficient to obtain the preprocessor.

[0012] Optionally, the closed-form solution is used to represent that the pixel value of the current pixel point is equal to the pixel value obtained in the previous iteration minus the gradient of the pixel value, and the gradient of the pixel value represents the product of the gradient function and the inverse matrix of the preprocessor.

[0013] According to another aspect of the present application, an image denoising device is provided. The device includes: an acquisition unit, configured to acquire a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function with the goal of minimizing the noise of the target image and an equality constraint condition of variables in the objective function; a determination unit, configured to determine an update rule for variables required to solve the composite convex minimization model, and determine a closed-form solution for the pixel value of a pixel point in the target image based on the required update rule for variables, where the closed-form solution includes second derivative information of the objective function; a solution unit, configured to iteratively solve the pixel value of each pixel point in the target image according to the closed-form solution to obtain the target pixel value of each pixel point in the case of minimizing the noise of the target image, and form a denoised image from the obtained target pixel values of each pixel point.

[0014] According to another aspect of the embodiments of the present invention, a computer storage medium is further provided. The computer storage medium includes a stored program, where when the program runs, it controls a device where the computer storage medium is located to execute an image denoising method.

[0015] According to another aspect of the embodiments of the present invention, an electronic device is further provided, including a processor and a memory; the memory stores computer-readable instructions, and the processor is configured to run the computer-readable instructions, where when the computer-readable instructions run, they execute an image denoising method.

[0016] Through this application, the following steps are adopted: obtaining a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function aiming at minimizing the noise of the target image and equality constraint conditions of variables in the objective function; determining an update rule for variables required to solve the composite convex minimization model, and determining a closed-form solution for the pixel values of pixel points in the target image based on the required update rule for variables, where the closed-form solution includes second-order derivative information of the objective function; iteratively solving the pixel values of each pixel point in the target image according to the closed-form solution to obtain the target pixel values of each pixel point in the case of minimizing the noise of the target image, and forming a denoised image from the obtained target pixel values of each pixel point, thus solving the problem of large computational complexity in accurately solving the pixel values of the denoised image when image denoising is a composite convex minimization problem in the related art. By determining a closed-form solution including second-order derivative information of the objective function and using this closed-form solution to solve the target pixel values of pixel points, the effect of reducing the computational complexity of solving the pixel values of the denoised image is achieved. Description of the Drawings

[0017] The drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0018] Figure 1 is a flowchart of an image denoising method provided according to an embodiment of this application;

[0019] Figure 2 is a schematic diagram of an image denoising device provided according to an embodiment of this application;

[0020] Figure 3 is a schematic diagram of an electronic device provided according to an embodiment of this application. Detailed Embodiments

[0021] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will describe this application in detail with reference to the drawings and in combination with the embodiments.

[0022] In order to enable those skilled in the art to better understand the solution of this application, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of this application.

[0023] It should be noted that in the description and claims of this application and the above-mentioned drawings, the terms "first", "second", etc. 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 as to implement the embodiments of this application 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 does not necessarily limit to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0024] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0025] According to an embodiment of the present application, an image denoising method is provided.

[0026] Figure 1 is a flowchart of the image denoising method according to an embodiment of the present application. As Figure 1 shown, the method includes the following steps:

[0027] Step S102, obtain a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function with the minimum noise of the target image as the goal and an equality constraint condition for the variables in the objective function.

[0028] Specifically, the target image is an image containing noise, and the composite convex minimization model for image denoising satisfies the form of, f(x)+g(y) is an objective function with the minimum noise of the target image as the goal, Ax + By = c is an equality constraint condition for the variables in the objective function, the variables in the objective function are the values of x and y, the objective function is composed of a loss function f(x) and a regularization term function g(y), and the loss function x refers to the denoised image, n is the number of pixel points in the denoised image, and x i refers to the value of the pixel point in the denoised image. Image denoising of the target image is to solve for x i when f(x)+g(y) is minimized.

[0029] Step S104, determine an update rule for the variables required to solve the composite convex minimization model, and determine a closed-form solution for the pixel values of the pixel points in the target image based on the required variable update rule, where the closed-form solution includes second-order derivative information of the objective function.

[0030] Specifically, the composite convex minimization model for image denoising satisfies The update rules for the required variables include the update rule for x, the update rule for y, and the update rules for other intermediate variables during the iterative solution process. The update rule for x is the update rule for each pixel of the target image. After obtaining the update rule for x, the update rule for x is then converted into a closed-form solution for x, that is, it is converted into an explicit expression that can directly represent the relationship between the pixel value of the current pixel and the pixel value obtained in the previous iteration.

[0031] Step S106: Iteratively solve the pixel values of each pixel in the target image according to the closed-form solution, obtain the target pixel values of each pixel in the case where the noise of the target image is minimized, and form the denoised image from the obtained target pixel values of each pixel.

[0032] Specifically, during the update of x, since the initial value assigned to x is known, and in the iterative process, the x obtained in the previous iteration can be obtained. Since the closed-form solution of x is an explicit expression that can directly represent the relationship between the pixel value of the current pixel and the pixel value obtained in the previous iteration, substituting the pixel value of the pixel obtained in the previous iteration into the formula of the closed-form solution can calculate the specific value, that is, the pixel value of the current iteration.

[0033] It should be noted that since the closed-form solution contains the second-order derivative information of the objective function, that is, the second-order gradient information in the gradient iterative update, introducing the second-order gradient information into the gradient iterative update significantly improves the convergence speed of the iteration and has a smaller gradient complexity under ill-conditioned conditions, thereby reducing the process of accurately calculating the pixel values of each pixel of the denoised image.

[0034] The image denoising method provided by the embodiments of the present application obtains a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function with the goal of minimizing the noise of the target image and an equality constraint condition for the variables in the objective function; determines the update rules for the variables required to solve the composite convex minimization model, and determines a closed-form solution for the pixel values of the pixels in the target image based on the update rules for the required variables, where the closed-form solution contains the second-order derivative information of the objective function; iteratively solves the pixel values of each pixel in the target image according to the closed-form solution, obtains the target pixel values of each pixel in the case where the noise of the target image is minimized, and forms the denoised image from the obtained target pixel values of each pixel, solving the problem of large computational complexity in accurately solving the pixel values of the denoised image in the related art when the image denoising is a composite convex minimization problem. By determining a closed-form solution containing the second-order derivative information of the objective function and using this closed-form solution to solve the target pixel values of the pixels, the effect of reducing the computational complexity of solving the pixel values of the denoised image is achieved.

[0035] Optionally, in the image denoising method provided in the embodiments of the present application, the objective function includes a loss function and a regularization term function. The variable of the loss function is the pixel value of the pixel points in the target image, and the variable of the regularization term function is the regularization term variable. Determining the update rules for the variables required to solve the composite convex minimization model includes: determining the augmented Lagrangian function according to the composite convex minimization model, where the variables of the augmented Lagrangian function include: the pixel value of the pixel points in the target image, the regularization term variable, and the dual variable; minimizing the augmented Lagrangian function with respect to the pixel value of the pixel points to obtain the update rule for the pixel value of the pixel points in the target image; minimizing the augmented Lagrangian function with respect to the regularization term variable to obtain the update rule for the regularization term variable; minimizing the augmented Lagrangian function with respect to the dual variable to obtain the update rule for the dual variable.

[0036] Specifically, in order to convert the solution problem of the composite convex minimization model for image denoising into an unconstrained solution problem, the augmented Lagrangian function can be determined according to the composite convex minimization model.

[0037] The composite convex minimization model for image denoising satisfies Converted to the augmented Lagrangian function: where ρ is the penalty parameter and u is the dual variable.

[0038] Further, minimizing the augmented Lagrangian function with respect to the pixel value of the pixel points to obtain the update rule for the pixel value of the pixel points in the target image, and then using the linearization method to convert the update rule for the pixel value of the pixel points in the target image into a closed-form solution: Minimizing the augmented Lagrangian function with respect to the regularization term variable, the update rule for the regularization term variable is: Minimizing the augmented Lagrangian function with respect to the dual variable, the update rule for the dual variable is: u t ←u t-1 +Ax t +By t -c.

[0039] Optionally, in the image denoising method provided in the embodiments of the present application, the closed-form solution is used to represent that the pixel value of the current pixel point is equal to the pixel value obtained in the previous iteration minus the gradient of the pixel value, and the gradient of the pixel value represents the product of the gradient function and the inverse matrix of the preprocessor.

[0040] Specifically, is the closed-form solution for x obtained in this embodiment, that is, the inverse matrix M of the preprocessor is left-multiplied in front of the gradient function -1 , where I tis a mini - batch sample of size b randomly drawn from {1, 2, …, n}, is the full gradient is an unbiased estimate, and its expected variance asymptotically approaches 0, so the learning rate η can be taken as a constant. In this embodiment, the pre - processor M is introduced into the iterative step of gradient descent through a closed - form solution. Since M contains the second - order information of the objective function, the speed of iterative solution can be accelerated.

[0041] Optionally, in the image denoising method provided in the embodiments of the present application, the closed - form solution of the pixel value of a pixel point in the target image is used to characterize the relationship between the pixel value of the current pixel point and the pixel value obtained in the previous iteration. By iteratively solving the pixel values of each pixel point in the target image according to the closed - form solution, the target pixel values of each pixel point in the case of the minimum noise in the target image are obtained, including: alternately iteratively solving the pixel value, the regularization term variable, and the dual variable according to the closed - form solution of the pixel value of the pixel point in the target image, the update rule of the regularization term variable, and the update rule of the dual variable until the preset number of iterations is reached, and obtaining the target pixel value of each pixel point.

[0042] Specifically, the code for alternating iterative solution is as follows:

[0043]

[0044] Among them, in the input parameters, m is the maximum number of inner - loop iterations, η is the step size (also called the learning rate), and ρ is the penalty parameter; are the initial values assigned to the original variables x, y, and the dual variable u. In the initial value of u represents finding the generalized inverse. The solution process is iteratively solved in two layers of inner and outer loops. In the inner loop, according to the update rules of x t , y t , u t iterate m times from t = 1, 2, …, m to update x t , y t iteratively, and obtain the output result

[0045] Optionally, in the image denoising method provided in the embodiments of the present application, the closed - form solution contains a pre - processor, and the pre - processor contains the second - order derivative information of the objective function.

[0046] Specifically, in , the pre - processor is represented by M. M can be a fixed pre - processor. For example, M is a symmetric positive - definite matrix, and M is a non - diagonal matrix, and the matrix contains the second - order derivative information of the objective function, so that the number of iterations can be less, the iterative process is significantly accelerated, and the gradient complexity is greatly reduced.

[0047] Optionally, in the image denoising method provided in the embodiments of the present application, the pre-processor is obtained in the following manner: taking the second derivative of the objective function to obtain the second derivative result; multiplying the second derivative result by a preset coefficient to obtain the pre-processor. That is, the M pre-processor can have a linear relationship with the second derivative of the objective function.

[0048] It should be noted that obtaining this pre-processor requires certain prerequisite conditions, that is, it is required that f i (x) has M -Lipschitz continuous gradient under ||·|| , f i (x) is M -strongly convex under ||·|| and the condition number of f i (x) under ||·|| M is

[0049] where f i (x) being M -strongly convex under ||·|| means that when there exists λ f > 0 such that for any x i , x j there is always then f i (x) is λ f -strongly convex. If there is always then f i (x) is M -strongly convex under ||·|| . Among them, for ||·|| M it is expressed as the M-induced norm, that is

[0050] f i (x) having M -Lipschitz continuous gradient under ||·|| means that when there exists L i > 0 such that for any x i , x j there is always then f i (x) has L i -Lipschitz continuous gradient. If there is always then f i (x) has M -Lipschitz continuous gradient under ||·|| .

[0051] f i(x) under ||·|| M The condition number under refers to: f i (x) under ||·|| M The condition number under is defined as When M = I (identity matrix), it is the condition number of f i (x).

[0052] Through this embodiment, a preprocessor containing second-order gradient information is introduced into the process of iteratively calculating the composite convex minimum problem, that is, a fixed preprocessor is used to accelerate the existing scheme. The preprocessor uses a non-diagonal positive definite matrix, so the number of iterations is less. On this basis, a closed-form solution of the optimization sub-problem is obtained, thus solving the problem that the actual performance deteriorates significantly under ill-conditioned conditions, improving the convergence speed and reducing the gradient complexity.

[0053] It should be noted that in addition to using a fixed preprocessor in this embodiment, other methods of introducing second-order gradient information can also be used, such as using a time-varying preprocessor. The form of the preprocessor is not limited in this embodiment.

[0054] Optionally, in the image denoising method provided in the embodiment of the present application, the objective function includes a loss function and a regularization term function. The regularization term function is determined by the norm of the product of the pixel value of the pixel point of the target image and a positive definite matrix, and the loss function is determined by the norm of the noise of the pixel point of the target image. The noise of the pixel point of the target image is determined by the difference between the pixel value of the pixel point and the pixel value of the pixel point of the original image.

[0055] Specifically, in the composite convex minimization model of image denoising in the embodiment of the present application, the loss function where the regularization term function g(x) = τ1||y||1. If B = I and C = 0 are set in the equality constraint Ax + By = c, then y = Ax and g(x) = τ1||Ax||1. Then the composite convex minimization model is where o i is the pixel value of the pixel point of the target image with Gaussian noise, x i is the pixel value of the pixel point of the original image, that is, the pixel value of the pixel point of the image after removing noise. The original image x i both follow the standard normal distribution and are independently and identically distributed. is a hyperparameter, τ1 is the coefficient of the first norm, A is the low-rank sparse representation of the target image, and the relationship between i and j can define the specific form of A. Specifically, when i = j, A ij = 1, when j = i + 1, A ij = -1, and in other cases A ij = 0. Solve The result is o i The pixel values of each pixel of the image in the case of the minimum noise, that is, x i value.

[0056] In addition, it should be noted that the method of this embodiment can be applied not only to the image denoising scenario, but also to other scenarios of composite convex minimization problems with equality constraint conditions, including fields such as large-scale machine learning, data science, statistics, and operations research. For example, it can be applied to the multi-task learning scenario; it can be applied to the graph-guided logistic regression problem; it can be applied to the graph-guided SVM problem; it can be applied to the regularized empirical risk minimization (ERM) problem; it can be applied to the ERM problem with a structured sparse regularization term. The application scenarios of this application are not limited.

[0057] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0058] The embodiment of the present application also provides an image denoising device. It should be noted that the image denoising device of the embodiment of the present application can be used to execute the image denoising method provided by the embodiment of the present application. The image denoising device provided by the embodiment of the present application is introduced below.

[0059] Figure 2 It is a schematic diagram of the image denoising device according to the embodiment of the present application. As Figure 2 shown, the device includes: an acquisition unit 21, a determination unit 22 and a solution unit 23.

[0060] Specifically, the acquisition unit 21 is configured to acquire a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function with the minimum noise of the target image as the objective and an equality constraint condition for variables in the objective function.

[0061] The determination unit 22 is configured to determine an update rule for variables required to solve the composite convex minimization model, and determine a closed-form solution for the pixel values of the pixels in the target image based on the required update rule for variables, where the closed-form solution includes second-order derivative information of the objective function.

[0062] The solution unit 23 is configured to iteratively solve the pixel values of each pixel in the target image according to the closed-form solution, obtain the target pixel values of each pixel in the case of the minimum noise of the target image, and form a denoised image from the obtained target pixel values of each pixel.

[0063] The image denoising device provided by the embodiment of the present application obtains, through an acquisition unit 21, a composite convex minimization model for performing image denoising on a target image. The composite convex minimization model includes an objective function aiming at minimizing the noise of the target image and an equality constraint condition of variables in the objective function. A determination unit 22 determines an update rule for variables required to solve the composite convex minimization model and determines a closed-form solution for the pixel values of pixel points in the target image based on the update rule of the required variables. The closed-form solution includes second-order derivative information of the objective function. A solution unit 23 iteratively solves the pixel values of each pixel point in the target image according to the closed-form solution, obtains the target pixel values of each pixel point in the case of minimizing the noise of the target image, and forms a denoised image from the obtained target pixel values of each pixel point, solving the problem of large computational complexity in accurately solving the pixel values of the denoised image when image denoising is a composite convex minimization problem in the related art. By determining a closed-form solution including second-order derivative information of the objective function and using the closed-form solution to solve the target pixel values of pixel points, the effect of reducing the computational complexity of solving the pixel values of the denoised image is achieved.

[0064] Optionally, in the image denoising device provided by the embodiment of the present application, the objective function includes a loss function and a regularization term function. The variable of the loss function is the pixel value of the pixel point in the target image, and the variable of the regularization term function is a regularization term variable. The determination unit 22 includes: a first determination module for determining an augmented Lagrangian function according to the composite convex minimization model, where the variables of the augmented Lagrangian function include: the pixel value of the pixel point in the target image, the regularization term variable, and the dual variable; a second determination module for minimizing the augmented Lagrangian function with respect to the pixel value of the pixel point to obtain an update rule for the pixel value of the pixel point in the target image; a third determination module for minimizing the augmented Lagrangian function with respect to the regularization term variable to obtain an update rule for the regularization term variable; a fourth determination module for minimizing the augmented Lagrangian function with respect to the dual variable to obtain an update rule for the dual variable.

[0065] Optionally, in the image denoising device provided by the embodiment of the present application, the closed-form solution for the pixel value of the pixel point in the target image is used to characterize the relationship between the pixel value of the current pixel point and the pixel value obtained in the previous iteration. The solution unit 23 includes: a first solution module for alternately iteratively solving the pixel value, the regularization term variable, and the dual variable according to the update rule of the pixel value of the pixel point in the target image, the update rule of the regularization term variable, and the update rule of the dual variable until the iteration preset number is reached; a second solution module for substituting each pixel value obtained after the iteration preset number into the closed-form solution to obtain the target pixel value of each pixel point.

[0066] Optionally, in the image denoising device provided in the embodiments of the present application, the objective function includes a loss function and a regularization term function. The regularization term function is determined by the norm of the product of the pixel values of the pixels of the target image and a positive definite matrix, and the loss function is determined by the norm of the noise of the pixels of the target image. The noise of the pixels of the target image is determined by the difference between the pixel values of the pixels and the pixel values of the pixels of the original image.

[0067] Optionally, in the image denoising device provided in the embodiments of the present application, the closed-form solution includes a pre-processor, and the pre-processor includes second-order derivative information of the objective function.

[0068] Optionally, in the image denoising device provided in the embodiments of the present application, the pre-processor is obtained by the following method: taking the second-order derivative of the objective function to obtain a second-order derivative result; multiplying the second-order derivative result by a preset coefficient to obtain the pre-processor.

[0069] Optionally, in the image denoising device provided in the embodiments of the present application, the closed-form solution is used to represent that the pixel value of the current pixel is equal to the pixel value obtained in the previous iteration minus the gradient of the pixel value, and the gradient of the pixel value represents the product of the gradient function and the inverse matrix of the pre-processor.

[0070] The above image denoising device includes a processor and a memory. The above acquisition unit 21, determination unit 22, solution unit 23, etc. are all stored in the memory as program units, and the corresponding functions are implemented by the processor executing the above program units stored in the memory.

[0071] The processor includes a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels can be set, and by adjusting the kernel parameters, the problem of large computational complexity in accurately solving the pixel values of the denoised image in the related art when image denoising is a composite convex minimization problem can be solved.

[0072] The memory may include non-permanent memory in a computer-readable medium, forms such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one storage chip.

[0073] The embodiments of the present application also provide a computer storage medium, and the computer storage medium includes a stored program. Wherein, when the program runs, it controls a device where the computer storage medium is located to execute an image denoising method.

[0074] The embodiments of the present application also provide an electronic device. Figure 3 is a schematic diagram of the electronic device provided in the embodiments of the present application, as Figure 3As shown, the electronic device 301 includes a processor and a memory; computer-readable instructions are stored in the memory, and the processor is used to run the computer-readable instructions. Among them, when the computer-readable instructions run, an image denoising method is executed. The electronic device in this article can be a server, a PC, a PAD, a mobile phone, etc.

[0075] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0076] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in one process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0077] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the specified functions in one process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0078] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Therefore, the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in one process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0079] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0080] The memory may include non - permanent memory in the form of computer - readable media, random access memory (RAM) and / or non - volatile memory such as read - only memory (ROM) or flash RAM. The memory is an example of computer - readable media.

[0081] Computer - readable media includes permanent and non - permanent, removable and non - removable media and can store information by any method or technology. The information can be computer - readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase - change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read - only memory (ROM), electrically erasable programmable read - only memory (EEPROM), flash memory or other memory technologies, compact disc read - only memory (CD - ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non - transitory media that can be used to store information accessible by a computing device. As defined herein, computer - readable media does not include transitory media such as modulated data signals and carrier waves.

[0082] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non - exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0083] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. An image denoising method, characterized in that, Including: Obtain a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function aiming at minimizing the noise of the target image and equality constraint conditions of variables in the objective function. The objective function includes a loss function and a regularization term function. The regularization term function is determined by the norm of the product of the pixel value of a pixel point in the target image and a positive definite matrix. The loss function is determined by the norm of the noise of a pixel point in the target image. The noise of a pixel point in the target image is determined by the difference between the pixel value of the pixel point and the pixel value of a pixel point in the original image, and the original image is an image with noise removed; Determine an update rule for variables required to solve the composite convex minimization model, and determine a closed-form solution for the pixel value of a pixel point in the target image based on the update rule of the required variables. The closed-form solution includes second-order derivative information of the objective function, and the closed-form solution for the pixel value of a pixel point in the target image is used to characterize the relationship between the pixel value of the current pixel point and the pixel value obtained in the previous iteration; Iteratively solve the pixel value of each pixel point in the target image according to the closed-form solution to obtain the target pixel value of each pixel point in the case of minimizing the noise of the target image, and form a denoised image from the obtained target pixel values of each pixel point.

2. The method according to claim 1, wherein Determining an update rule for variables required to solve the composite convex minimization model includes: Determine an augmented Lagrangian function according to the composite convex minimization model, where the variables of the augmented Lagrangian function include: the pixel value of a pixel point in the target image, a regularization term variable, and a dual variable; Minimize the augmented Lagrangian function with respect to the pixel value of a pixel point to obtain an update rule for the pixel value of a pixel point in the target image; Minimize the augmented Lagrangian function with respect to the regularization term variable to obtain an update rule for the regularization term variable; Minimize the augmented Lagrangian function with respect to the dual variable to obtain an update rule for the dual variable.

3. The method according to claim 2, wherein The variable of the loss function is the pixel value of a pixel point in the target image, and the variable of the regularization term function is the regularization term variable. Iteratively solving the pixel value of each pixel point in the target image according to the closed-form solution to obtain the target pixel value of each pixel point in the case of minimizing the noise of the target image includes: Iteratively solve the pixel value, the regularization term variable, and the dual variable alternately according to the closed-form solution of the pixel point in the target image, the update rule of the regularization term variable, and the update rule of the dual variable until the iteration preset number is reached to obtain the target pixel value of each pixel point.

4. The method according to claim 1, characterized in that, The closed-form solution includes a preprocessor, and the preprocessor includes second-order derivative information of the objective function.

5. The method according to claim 4, wherein The preprocessor is obtained by the following method: Take the second derivative of the objective function to obtain a second derivative result; Multiply the second derivative result by a preset coefficient to obtain the preprocessor.

6. The method according to claim 4, wherein The closed-form solution is used to represent that the pixel value of the current pixel point is equal to the pixel value obtained in the previous iteration minus the gradient of the pixel value, and the gradient of the pixel value represents the product of the gradient function and the inverse matrix of the preprocessor.

7. An image denoising device, characterized in that, It includes: An acquisition unit, configured to acquire a composite convex minimization model for image denoising of a target image, where the composite convex minimization model includes an objective function aiming at minimizing the noise of the target image, an equality constraint condition of variables in the objective function, the objective function includes a loss function and a regularization term function, the regularization term function is determined by the norm of the product of the pixel value of the pixel point of the target image and a positive definite matrix, the loss function is determined by the norm of the noise of the pixel point of the target image, the noise of the pixel point of the target image is determined by the difference between the pixel value of the pixel point and the pixel value of the pixel point of the original image, and the original image is an image from which noise has been removed; A determination unit, configured to determine an update rule for variables required to solve the composite convex minimization model, and determine a closed-form solution of the pixel value of the pixel point in the target image based on the update rule for the required variables, where the closed-form solution includes second-order derivative information of the objective function, and the closed-form solution of the pixel value of the pixel point in the target image is used to represent the relationship between the pixel value of the current pixel point and the pixel value obtained in the previous iteration; A solution unit, configured to iteratively solve the pixel values of each pixel point in the target image according to the closed-form solution, obtain the target pixel values of each pixel point in the case of minimizing the noise of the target image, and form a denoised image from the obtained target pixel values of each pixel point.

8. A computer storage medium, characterized in that, The computer storage medium includes a stored program, where the program, when running, controls the device where the computer storage medium is located to execute the image denoising method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, It includes a processor and a memory, where computer-readable instructions are stored in the memory, and the processor is configured to run the computer-readable instructions, where the computer-readable instructions, when running, execute the image denoising method according to any one of claims 1 to 6.

Images