Image noise removal system, method, device and storage medium

Abstract

The present invention discloses an image noise removal system, method, device, and storage medium, including a host computer and a slave computer. The slave computer includes at least one device, each of which is deployed with a corresponding image denoising algorithm. The host computer is used to generate at least one new image denoising algorithm parameter based on the initial image texture loss value and initial image denoising algorithm parameters corresponding to the original noisy image in each device obtained. The slave computer is used to update the image denoising algorithm according to each new image denoising algorithm parameter and perform denoising processing on the original noisy image according to the updated image denoising algorithm. Through communication interaction between the host computer and the slave computer, the present invention can simultaneously debug and update multiple image denoising algorithms deployed in the slave computer, reducing debugging costs, improving debugging efficiency and accuracy, avoiding environmental constraints, and realizing denoising processing of the original noisy image according to the updated image denoising algorithm, thereby improving image noise removal efficiency.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an image noise removal system, method, device and storage medium. Background Art

[0002] During the generation and transmission process, images are often interfered with and affected by various noises, which leads to image degradation and has an adverse impact on subsequent image processing and image visual effects.

[0003] With the continuous development of cameras, major camera equipment manufacturers have increasingly higher requirements for the quality of photos taken, and image noise removal is an important factor affecting the final image quality. Currently, most major camera manufacturers rely on manual debugging to debug the effect of image noise removal. This process often requires manual modification of parameters, shooting, comparison, etc., which consumes a lot of time and manpower costs, and the accuracy is difficult to control. In addition, manual debugging can only be done by one person at a time for one device corresponding to one set of image denoising algorithms. For multi-device platforms or devices that integrate multiple sets of image denoising algorithms, manual debugging obviously requires more manpower and more time costs.

[0004] At the same time, due to differences in hardware, image denoising algorithms, etc. among camera device manufacturers, the environment for image noise removal debugging is variable. Manual debugging is restricted by the environment, and it is impossible to adopt a unified debugging strategy for different hardware and different image denoising algorithms. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of the existing technology of removing image noise by manual debugging, such as the inability to debug multiple sets of image denoising algorithms at the same time, low debugging accuracy, and being restricted by the environment. An image noise removal system, method, device and storage medium are provided.

[0006] The present invention solves the above technical problems through the following technical solutions:

[0007] A first aspect of the present invention provides an image noise removal system, comprising a host computer and a slave computer, wherein the host computer is communicatively connected to the slave computer, and the slave computer comprises at least one device, each device being deployed with a corresponding image denoising algorithm;

[0008] The upper computer is used to obtain the initial image texture loss value and the initial image denoising algorithm parameter corresponding to the original noisy image in each device from the lower computer, so as to generate at least one new image denoising algorithm parameter according to each of the initial image texture loss value and the initial image denoising algorithm parameter, and send the at least one new image denoising algorithm parameter to the lower computer;

[0009] The lower computer is used to update the image denoising algorithm in each corresponding device according to each new image denoising algorithm parameter, and perform denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image.

[0010] Preferably, the host computer includes a host computer communication module and an asynchronous policy gradient algorithm module, and the slave computer includes an image denoising processing module;

[0011] The upper computer communication module is used to obtain the initial image texture loss value and the initial image denoising algorithm parameters corresponding to the original noisy image in each device from the lower computer, and send each of the initial image texture loss value and the initial image denoising algorithm parameters to the asynchronous policy gradient algorithm module;

[0012] The asynchronous policy gradient algorithm module is used to generate at least one new image denoising algorithm parameter according to each of the initial image texture loss values and the initial image denoising algorithm parameters, and send the at least one new image denoising algorithm parameter to the image denoising processing module through the host computer communication module;

[0013] The image denoising processing module is used to update the image denoising algorithm in each corresponding device according to each of the new image denoising algorithm parameters, and perform denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image.

[0014] Preferably, the lower computer further includes an image texture loss calculation module and a lower computer communication module;

[0015] The asynchronous policy gradient algorithm module is further configured to update the network parameters of the asynchronous policy gradient algorithm module upon receiving the initial image texture loss value and the initial image denoising algorithm parameters, and send the updated iteration rounds to the image texture loss calculation module through the host computer communication module;

[0016] The image texture loss calculation module is used to calculate a new image texture loss value corresponding to the denoised image, and send the new image texture loss value to the lower computer communication module;

[0017] The image texture loss calculation module is further configured to set the lower computer to a training completion state and call the image denoising processing module when the new image texture loss value converges and / or the update iteration round reaches a preset threshold;

[0018] The image denoising processing module is further configured to output a target denoised image.

[0019] Preferably, the lower computer communication module is used to send the new image texture loss value to the asynchronous policy gradient algorithm module when the new image texture loss value has not converged;

[0020] The asynchronous policy gradient algorithm module is used to re-update the network parameters of the asynchronous policy gradient algorithm module and regenerate new image denoising algorithm parameters.

[0021] Preferably, the lower computer communication module is used to send the new image texture loss value to the asynchronous policy gradient algorithm module when the update iteration round does not reach a preset threshold;

[0022] The asynchronous policy gradient algorithm module is used to re-update the network parameters of the asynchronous policy gradient algorithm module according to the new image texture loss value and regenerate new image denoising algorithm parameters.

[0023] Preferably, the host computer further includes a device management module, the asynchronous policy gradient algorithm module includes a main control network and several sub-networks, each sub-network has a corresponding sub-thread, and each sub-thread interacts with a corresponding device;

[0024] The device management module is used to obtain the operating status of each device in the lower computer and send the operating status of each device to the asynchronous policy gradient algorithm module;

[0025] The main control network is used to control the opening or closing of the corresponding sub-network according to the operating status of each device.

[0026] A second aspect of the present invention provides an image noise removal method, which is implemented using the image noise removal system described in the first aspect. The image noise removal method includes:

[0027] Obtain the initial image texture loss value and initial image denoising algorithm parameters corresponding to the original noisy image in each device;

[0028] generating at least one new image denoising algorithm parameter according to each of the initial image texture loss values and the initial image denoising algorithm parameter;

[0029] updating the image denoising algorithm in each of the corresponding devices according to each of the new image denoising algorithm parameters;

[0030] The original noisy image corresponding to the updated image denoising algorithm is denoised according to the updated image denoising algorithm to obtain a denoised image.

[0031] Preferably, before the step of obtaining the initial image texture loss value and initial image denoising algorithm parameters corresponding to the original noisy image in each device, the image noise removal method further comprises:

[0032] Start the host computer;

[0033] Determine whether the upper computer is in an idle state, and if so, start the lower computer and set the lower computer to a working state;

[0034] Initialize the network parameters of the asynchronous policy gradient algorithm module;

[0035] The step of generating at least one new image denoising algorithm parameter according to each of the initial image texture loss values and the initial image denoising algorithm parameter comprises:

[0036] Update the network parameters of the asynchronous policy gradient algorithm module according to each of the initial image texture loss values and the initial image denoising algorithm parameters and generate corresponding new image denoising algorithm parameters;

[0037] After the step of performing denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image, the image noise removal method further includes:

[0038] Calculate the new image texture loss value corresponding to the denoised image;

[0039] Determine whether the new image texture loss value converges and / or whether the update iteration round reaches a preset threshold, and if so, set the lower computer to a completed training state;

[0040] Output the target denoised image.

[0041] A third aspect of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the image noise removal method described in the second aspect when executing the computer program.

[0042] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the image noise removal method described in the second aspect.

[0043] On the basis of conforming to the common sense in this field, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present invention.

[0044] The positive progress effect of the present invention is:

[0045] Through communication interaction between the upper computer and the lower computer, the present invention can simultaneously debug and update multiple sets of image denoising algorithms deployed in the lower computer. Compared with the existing manual debugging method in which one person can only debug and update one set of image denoising algorithms corresponding to one device at a time, the debugging cost is lower, the debugging efficiency is higher and more accurate, and the environmental constraints are avoided. The original noisy image corresponding to each set of updated image denoising algorithms is denoised according to the updated image denoising algorithms, thereby improving the image noise removal efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 FIG. 4 is a structural diagram of an image noise removal system according to embodiment 1 of the present invention.

[0047] Figure 2 FIG. 4 is another structural diagram of the image noise removal system according to Example 1 of the present invention.

[0048] Figure 3 This is a structural diagram of the asynchronous policy gradient algorithm module of the image noise removal system in Example 1 of the present invention.

[0049] Figure 4 This is a first flow chart of the image noise removal method according to embodiment 2 of the present invention.

[0050] Figure 5 This is a second flow chart of the image noise removal method according to embodiment 2 of the present invention.

[0051] Figure 6 This is a schematic structural diagram of an electronic device according to embodiment 3 of the present invention. DETAILED DESCRIPTION

[0052] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0053] Example 1

[0054] like Figure 1 As shown, the image noise removal system of this embodiment includes a host computer 1 and a slave computer 2, the host computer 1 is in communication with the slave computer 2, and the slave computer 2 includes at least one device N (N=1, 2...n), each of which is deployed with a corresponding image denoising algorithm;

[0055] In this embodiment, each device is integrated with a dedicated image denoising algorithm IP. The device may be an imaging device or a photographing device, or other devices, which are not specifically limited here.

[0056] The upper computer 1 is used to obtain the initial image texture loss value and the initial image denoising algorithm parameters corresponding to the original noisy image in each device from the lower computer 2, generate at least one new image denoising algorithm parameter according to each initial image texture loss value and the initial image denoising algorithm parameter, and send the at least one new image denoising algorithm parameter to the lower computer 2;

[0057] In this embodiment, for the multiple original noise images, the multiple original noise images may be the same image or different images, which is not specifically limited here.

[0058] The lower computer 2 is used to update the image denoising algorithm in each corresponding device according to each new image denoising algorithm parameter, and perform denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image.

[0059] This embodiment adopts a division of labor and coordination between the upper and lower computers. The upper computer, as an intelligent agent, is responsible for making decisions on each image denoising algorithm parameter update and sending new image denoising algorithm parameters to the lower computer; the lower computer, as a carrier of the environmental state, is responsible for image denoising processing after each image denoising algorithm parameter update and sending the texture loss value of the new denoised image to the upper computer. Compared with the existing manual debugging method, the image denoising algorithm parameters are automatically adjusted through computer learning and decision-making until the best image denoising effect is achieved. At the same time, it can also adapt to different debugging environments, such as different hardware and different image denoising algorithms, avoiding the constraints of the debugging environment, and can debug the denoising effects of multiple devices at the same time, saving more time and labor costs.

[0060] In one feasible solution, Figure 2 As shown, the host computer 1 includes a host computer communication module 11 and an asynchronous policy gradient algorithm module 12, and the lower computer 2 includes an image denoising processing module 21;

[0061] The upper computer communication module 11 is used to obtain the initial image texture loss value and the initial image denoising algorithm parameters corresponding to the original noisy image in each device from the lower computer 2, and send each initial image texture loss value and the initial image denoising algorithm parameters to the asynchronous policy gradient algorithm module 12; specifically, the upper computer communication module 11 is used to obtain the initial image texture loss value and the initial image denoising algorithm parameters corresponding to the original noisy image in each device from the lower computer 2, and obtain the real-time image texture loss value of each device in the lower computer 2 in each round of iteration, and send each initial image texture loss value, the initial image denoising algorithm parameters and the real-time image texture loss value corresponding to each round of iteration to the asynchronous policy gradient algorithm module 12;

[0062] In this embodiment, the host computer communication module 11 is mainly used to control the communication interaction between the host computer 1 and the slave computer 2. The host computer communication module 11 is provided with an image denoising algorithm parameter sending interface, a texture loss value acquisition interface and a device status acquisition interface.

[0063] The asynchronous policy gradient algorithm module 12 is used to generate at least one new image denoising algorithm parameter according to each initial image texture loss value and the initial image denoising algorithm parameter, and send the at least one new image denoising algorithm parameter to the image denoising processing module 21 through the host computer communication module 11;

[0064] In this embodiment, each initial image texture loss value and initial image denoising algorithm parameters are input into the asynchronous policy gradient algorithm module 12 to output at least one new image denoising algorithm parameter.

[0065] The image denoising processing module 21 is used to update the image denoising algorithm in each corresponding device according to each new image denoising algorithm parameter, and perform denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image.

[0066] This embodiment adopts an asynchronous policy gradient algorithm module, in which the computer automatically adjusts the image denoising algorithm parameters according to changes in the debugging environment, eliminating the need for manual debugging and improving debugging efficiency and accuracy.

[0067] In one feasible solution, Figure 2 As shown, the lower computer 2 includes an image texture loss calculation module 22 and a lower computer communication module 23;

[0068] The asynchronous policy gradient algorithm module 12 is further configured to update the network parameters of the asynchronous policy gradient algorithm module upon receiving the initial image texture loss value and the initial image denoising algorithm parameters, and send the updated iteration rounds to the image texture loss calculation module 22 via the host computer communication module 11;

[0069] The image texture loss calculation module 22 is used to calculate the new image texture loss value corresponding to the denoised image and send the new image texture loss value to the lower computer communication module 23;

[0070] The image texture loss calculation module 22 is further configured to set the lower computer 2 to a training completion state and call the image denoising processing module 21 when the new image texture loss value converges and / or the update iteration rounds reach a preset threshold;

[0071] The called image denoising processing module 21 is further configured to output a target denoised image.

[0072] Specifically, when the new image texture loss value converges and / or the update iteration rounds reach a preset threshold, the image texture loss calculation module 22 sets the lower computer 2 to the completed training state and outputs the target denoised image through the image denoising processing module 21.

[0073] In this embodiment, the lower computer communication module 23 is mainly used to control the communication interaction between the lower computer 2 and the upper computer 1. The lower computer communication module 23 is provided with an image denoising algorithm parameter receiving interface, a texture loss value sending interface and a device status sending interface.

[0074] In this embodiment, when the new image texture loss value tends to a stable value, it is determined that the new image texture loss value converges.

[0075] It should be noted that the stable value and the preset threshold are set according to actual conditions and are not specifically limited here.

[0076] In an implementable solution, the lower computer communication module 23 is used to send the new image texture loss value to the asynchronous policy gradient algorithm module 12 when the new image texture loss value has not converged;

[0077] The asynchronous policy gradient algorithm module 12 is used to re-update the network parameters of the asynchronous policy gradient algorithm module and regenerate new image denoising algorithm parameters.

[0078] In an implementable solution, the lower computer communication module 23 is used to send the new image texture loss value to the asynchronous policy gradient algorithm module 12 when the update iteration round does not reach the preset threshold;

[0079] The asynchronous policy gradient algorithm module 12 is used to re-update the network parameters of the asynchronous policy gradient algorithm module according to the new image texture loss value and regenerate new image denoising algorithm parameters.

[0080] In one feasible solution, Figure 2 As shown, the host computer 1 also includes a device management module 13, such as Figure 3 As shown, the asynchronous policy gradient algorithm module 12 includes a main control network and several sub-networks, each sub-network has a corresponding sub-thread, and each sub-thread interacts with a corresponding device;

[0081] The device management module 13 is used to obtain the operating status of each device in the lower computer 2 and send the operating status of each device to the asynchronous policy gradient algorithm module 12;

[0082] The master network is used to control the opening or closing of the corresponding sub-network according to the operating status of each device.

[0083] In this embodiment, the operating status of each device includes but is not limited to IDLE (idle state), WORK (working state), DONE (training completed state), and UNUSUAL (abnormal state);

[0084] In the specific implementation process, the implementation principle of the asynchronous policy gradient algorithm module 12 is as follows Figure 3 As shown:

[0085] (1) Realize the master control network;

[0086] In this embodiment, the main control network is mainly used to control the opening or closing of each sub-network, and the opening or closing is determined by the operating status of the device corresponding to the sub-network.

[0087] (2) Implement sub-networks;

[0088] In this embodiment, to implement the sub-network in (2), N sub-threads need to be established, where N is the number of devices in the lower computer. Each sub-thread is responsible for interacting with one lower computer device. At the same time, an asynchronous policy gradient model is deployed in each sub-thread. The asynchronous policy gradient model is fitted by a BP neural network. The input of the BP neural network is the current image denoising algorithm parameter S, and the output is the new image denoising algorithm parameter S'.

[0089] It should be noted that the device management module 13 is mainly used to obtain the operating status of each device in the current lower computer. After obtaining the operating status of the device, the device management module 13 notifies the master control network in (1) to control the sub-network. The specific control process is as follows:

[0090] When the device is in IDLE state, the master network opens the sub-network corresponding to the device;

[0091] When the device is in DONE or UNUSUAL state, the master network closes the sub-network corresponding to the device;

[0092] When the device is in the WORK state, the master network does not perform any operations.

[0093] In this embodiment, an asynchronous multi-threaded network structure is adopted, and a host computer can communicate and interact with multiple devices in the slave computer at the same time. Each sub-thread controls one device, which effectively utilizes computer resources and improves the efficiency of automatic debugging. In addition, the initial debugging environment corresponding to each device is configurable, and the parameters of each sub-network are not correlated with each other, which improves the anti-interference ability of the image noise removal system and enables the image noise removal system to adapt to different debugging environments.

[0094] It should be noted that the image noise removal system in this embodiment may be a separate chip, a chip module or an image noise removal device, or may be a chip or chip module integrated into the image noise removal device.

[0095] This embodiment can simultaneously debug and update multiple sets of image denoising algorithms deployed in the lower computer through communication interaction between the upper computer and the lower computer. Compared with the existing manual debugging method in which one person can only debug and update one set of image denoising algorithms corresponding to one device at a time, the debugging cost is lower, the debugging efficiency is higher and more accurate, and the environmental constraints are avoided. It realizes denoising processing of the original noisy image corresponding to each set of updated image denoising algorithms according to each set of updated image denoising algorithms, thereby improving the efficiency of image noise removal.

[0096] Example 2

[0097] The image noise removal method of this embodiment is implemented using the image noise removal system of Example 1 (for example, the image noise removal method of this embodiment can be implemented using a separate chip, chip module, or image noise removal device, or can be implemented using a chip or chip module integrated into the image noise removal device). Figure 4 As shown, the image noise removal method includes:

[0098] Step 101: Obtain an initial image texture loss value and initial image denoising algorithm parameters corresponding to the original noisy image in each device;

[0099] In this embodiment, each device integrates a dedicated image denoising algorithm IP.

[0100] In this embodiment, for the multiple original noise images, the multiple original noise images may be the same image or different images, which is not specifically limited here.

[0101] Step 102: generating at least one new image denoising algorithm parameter according to each initial image texture loss value and the initial image denoising algorithm parameter;

[0102] Step 103: updating the image denoising algorithm in each corresponding device according to each new image denoising algorithm parameter;

[0103] Step 104 : Denoising the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image.

[0104] This embodiment adopts a division of labor and coordination between the upper and lower computers. The upper computer, as an intelligent agent, is responsible for making decisions on each image denoising algorithm parameter update and sending new image denoising algorithm parameters to the lower computer; the lower computer, as a carrier of the environmental state, is responsible for image denoising processing after each image denoising algorithm parameter update and sending the texture loss value of the new denoised image to the upper computer. Compared with the existing manual debugging method, the image denoising algorithm parameters are automatically adjusted through computer learning and decision-making until the best image denoising effect is achieved. At the same time, it can also adapt to different debugging environments, such as different hardware and different image denoising algorithms, avoiding the constraints of the debugging environment, and can debug the denoising effects of multiple devices at the same time, saving more time and labor costs.

[0105] In one feasible solution, Figure 5 As shown, before step 101, the image noise removal method further includes:

[0106] Step 100: Start the host computer;

[0107] Step 100-1: Determine whether the host computer is in an idle state. If so, execute step 100-2; if not, execute step 100-4;

[0108] Step 100-2: Start the slave computer and set it to working state;

[0109] Step 100-3: Initialize the network parameters of the asynchronous policy gradient algorithm module;

[0110] Step 100-4: Restart the slave computer and set it to idle state, then execute step 100-2;

[0111] Step 102 includes:

[0112] Step 102', updating the network parameters of the asynchronous policy gradient algorithm module according to each initial image texture loss value and the initial image denoising algorithm parameters and generating corresponding new image denoising algorithm parameters;

[0113] This embodiment adopts an asynchronous policy gradient algorithm module, in which the computer automatically adjusts the image denoising algorithm parameters according to changes in the debugging environment, eliminating the need for manual debugging and improving debugging efficiency and accuracy.

[0114] like Figure 5 As shown, after step 104, the image noise removal method further includes:

[0115] Step 105: Calculate a new image texture loss value corresponding to the denoised image;

[0116] Step 106: Determine whether the new image texture loss value converges and / or whether the update iteration round reaches a preset threshold. If so, execute step 107; if not, execute step 109;

[0117] In this embodiment, when the new image texture loss value tends to a stable value, it is determined that the new image texture loss value converges.

[0118] It should be noted that the stable value and the preset threshold are set according to actual conditions and are not specifically limited here.

[0119] Step 107: Set the lower computer to a training completion state;

[0120] Step 108: Output the target denoised image;

[0121] Step 109: Upload the new image texture loss value, and then execute step 102'.

[0122] This embodiment can simultaneously debug and update multiple sets of image denoising algorithms deployed in the lower computer through communication interaction between the upper computer and the lower computer. Compared with the existing manual debugging method in which one person can only debug and update one set of image denoising algorithms corresponding to one device at a time, the debugging cost is lower, the debugging efficiency is higher and more accurate, and the environmental constraints are avoided. It realizes denoising processing of the original noisy image corresponding to each set of updated image denoising algorithms according to each set of updated image denoising algorithms, thereby improving the efficiency of image noise removal.

[0123] Example 3

[0124] Figure 6 This is a schematic diagram of the structure of an electronic device provided in Example 3 of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the image noise removal method of Example 2 is implemented. Figure 6 The electronic device 30 shown is only an example and should not limit the functionality and scope of use of the embodiments of the present invention.

[0125] like Figure 6 As shown, the electronic device 30 may be a general-purpose computing device, such as a server device. Components of the electronic device 30 may include, but are not limited to, the at least one processor 31, the at least one memory 32, and a bus 33 connecting different system components (including the memory 32 and the processor 31).

[0126] The bus 33 includes a data bus, an address bus, and a control bus.

[0127] The memory 32 may include a volatile memory, such as a random access memory (RAM) 321 and / or a cache memory 322 , and may further include a read-only memory (ROM) 323 .

[0128] The memory 32 may also include a program / utility 325 having a set (at least one) of program modules 324, such program modules 324 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0129] The processor 31 executes various functional applications and data processing by running computer programs stored in the memory 32 , such as the image noise removal method of the second embodiment of the present invention.

[0130] The electronic device 30 may also communicate with one or more external devices 34 (e.g., a keyboard, a pointing device, etc.). Such communication may be performed via an input / output (I / O) interface 35. Furthermore, the model generating device 30 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 36. Figure 6 As shown, the network adapter 36 communicates with the other modules of the model-generated device 30 via the bus 33. It should be understood that, although not shown in the figures, other hardware and / or software modules may be used in conjunction with the model-generated device 30, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0131] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more units / modules described above may be embodied in a single unit / module. Conversely, the features and functions of a single unit / module described above may be further divided and embodied by multiple units / modules.

[0132] Example 4

[0133] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the image noise removal method provided in Embodiment 2 is implemented.

[0134] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0135] In a possible implementation manner, the present invention may also be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the image noise removal method described in Example 2.

[0136] The program code for executing the present invention may be written in any combination of one or more programming languages, and may be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0137] Although specific embodiments of the present invention have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, and such changes and modifications are intended to fall within the scope of the present invention.

Claims

1. An image noise removal system, characterized in that: The system comprises a host computer and a slave computer, wherein the host computer is in communication with the slave computer, and the slave computer comprises at least one device, each of which is equipped with a corresponding image denoising algorithm; The upper computer is used to obtain the initial image texture loss value and the initial image denoising algorithm parameter corresponding to the original noisy image in each device from the lower computer, so as to generate at least one new image denoising algorithm parameter according to each of the initial image texture loss value and the initial image denoising algorithm parameter, and send the at least one new image denoising algorithm parameter to the lower computer; The lower computer is used to update the image denoising algorithm in each of the corresponding devices according to each of the new image denoising algorithm parameters, and perform denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image; The host computer includes an asynchronous policy gradient algorithm module, and the slave computer includes an image denoising processing module, an image texture loss calculation module and a slave computer communication module; The asynchronous policy gradient algorithm module is further configured to update the network parameters of the asynchronous policy gradient algorithm module upon receiving the initial image texture loss value and the initial image denoising algorithm parameters, and send the updated iteration rounds to the image texture loss calculation module through the host computer communication module; The image texture loss calculation module is used to calculate a new image texture loss value corresponding to the denoised image, and send the new image texture loss value to the lower computer communication module; The image texture loss calculation module is further configured to set the lower computer to a training completion state and call the image denoising processing module when the new image texture loss value converges and / or the update iteration round reaches a preset threshold; The image denoising processing module is further configured to output a target denoised image; The lower computer communication module is used to send the new image texture loss value to the asynchronous policy gradient algorithm module to regenerate new image denoising algorithm parameters when the new image texture loss value has not converged and / or the update iteration round has not reached a preset threshold.

2. The image noise removal system according to claim 1, wherein: The host computer includes a host computer communication module and an asynchronous policy gradient algorithm module, and the slave computer includes an image denoising processing module; The upper computer communication module is used to obtain the initial image texture loss value and the initial image denoising algorithm parameters corresponding to the original noisy image in each device from the lower computer, and send each of the initial image texture loss value and the initial image denoising algorithm parameters to the asynchronous policy gradient algorithm module; The asynchronous policy gradient algorithm module is used to generate at least one new image denoising algorithm parameter according to each of the initial image texture loss values and the initial image denoising algorithm parameters, and send the at least one new image denoising algorithm parameter to the image denoising processing module through the host computer communication module; The image denoising processing module is used to update the image denoising algorithm in each corresponding device according to each of the new image denoising algorithm parameters, and perform denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image.

3. The image noise removal system according to claim 1, wherein: The lower computer communication module is used to send the new image texture loss value to the asynchronous policy gradient algorithm module when the new image texture loss value has not converged; The asynchronous policy gradient algorithm module is used to re-update the network parameters of the asynchronous policy gradient algorithm module and regenerate new image denoising algorithm parameters.

4. The image noise removal system according to claim 1, wherein: The lower computer communication module is used to send the new image texture loss value to the asynchronous policy gradient algorithm module when the update iteration round does not reach a preset threshold; The asynchronous policy gradient algorithm module is used to re-update the network parameters of the asynchronous policy gradient algorithm module according to the new image texture loss value and regenerate new image denoising algorithm parameters.

5. The image noise removal system according to claim 2, wherein: The host computer also includes a device management module, and the asynchronous policy gradient algorithm module includes a main control network and several sub-networks, each sub-network has a corresponding sub-thread, and each sub-thread interacts with a corresponding device; The device management module is used to obtain the operating status of each device in the lower computer and send the operating status of each device to the asynchronous policy gradient algorithm module; The main control network is used to control the opening or closing of the corresponding sub-network according to the operating status of each device.

6. A method for removing image noise, characterized in that: The image noise removal method is implemented using the image noise removal system according to any one of claims 1 to 5, and the image noise removal method includes: Obtain the initial image texture loss value and initial image denoising algorithm parameters corresponding to the original noisy image in each device; generating at least one new image denoising algorithm parameter according to each of the initial image texture loss values and the initial image denoising algorithm parameter; updating the image denoising algorithm in each of the corresponding devices according to each of the new image denoising algorithm parameters; performing denoising processing on the original noisy image corresponding to the updated image denoising algorithm according to the updated image denoising algorithm to obtain a denoised image; The image noise removal method also includes: calculating a new image texture loss value corresponding to the denoised image; judging whether the new image texture loss value converges and / or whether the update iteration round reaches a preset threshold; if so, setting the lower computer to a completed training state; outputting the target denoised image; if not, uploading the new image texture loss value, updating the network parameters of the asynchronous policy gradient algorithm module according to each of the initial image texture loss values and the initial image denoising algorithm parameters, and generating corresponding new image denoising algorithm parameters.

7. The image noise removal method according to claim 6, wherein: Before the step of obtaining the initial image texture loss value and initial image denoising algorithm parameters corresponding to the original noisy image in each device, the image noise removal method further includes: Start the host computer; Determine whether the upper computer is in an idle state, and if so, start the lower computer and set the lower computer to a working state; Initialize the network parameters of the asynchronous policy gradient algorithm module; The step of generating at least one new image denoising algorithm parameter according to each of the initial image texture loss values and the initial image denoising algorithm parameter comprises: Update the network parameters of the asynchronous policy gradient algorithm module according to each of the initial image texture loss values and the initial image denoising algorithm parameters and generate corresponding new image denoising algorithm parameters; The updated image denoising algorithm is used to perform denoising processing on the original noisy image corresponding to the updated image denoising algorithm to obtain a denoised image.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the image noise removal method according to any one of claims 6 to 7 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the image noise removal method according to any one of claims 6 to 7 is implemented.

Images