Data processing method and device

Abstract

The invention discloses a data processing method and device, and the method comprises the steps: obtaining the compression result data of a to-be-processed video; the compression result data comprises coded data of each video frame of the to-be-processed video and a target number of iterations corresponding to each video frame; the target number of iterations is the number of iterations capable of enabling the image quality corresponding to the denoising result to meet the image quality condition when the decoding data corresponding to the coded data is subjected to iterative denoising by using a pre-constructed loop filter; decoding coded data of a video frame in the compression result data to obtain corresponding decoded data; performing iterative denoising on the decoded data by using the loop filter to obtain a denoising result; wherein the number of iterations is the target number of iterations; and outputting a picture corresponding to a denoising result obtained by executing iterative denoising.

Description

Technical Field

[0001] This application belongs to the technical fields of image processing and artificial intelligence, and particularly relates to a data processing method and apparatus. Background Art

[0002] In order to improve the compression ratio, lossy compression methods are usually required to highly compress videos (such as 1000 - 10000 times). Lossy compression methods discard high-frequency signals that are insensitive to the human eye during the compression encoding process. Although discarding high-frequency signals can reduce the amount of signals to be encoded and increase the compression ratio, it will also introduce distortion and cause a decline in image quality. Therefore, how to improve the image quality on the premise of ensuring the required compression ratio has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0003] For this reason, this application discloses the following technical solutions:

[0004] A data processing method includes:

[0005] Obtaining compressed result data of a video to be processed; the compressed result data includes encoded data of each video frame of the video to be processed and target iteration times respectively corresponding to each video frame; the target iteration time is: when iteratively denoising the decoded data corresponding to the encoded data by using a pre-constructed loop filter, the iteration time that can make the image quality corresponding to the denoising result meet the image quality condition;

[0006] Decoding the encoded data of the video frame in the compressed result data to obtain corresponding decoded data;

[0007] Performing iterative denoising on the decoded data by using the loop filter to obtain a denoising result; wherein, the iteration time is the target iteration time;

[0008] Outputting a picture corresponding to the denoising result obtained by performing iterative denoising;

[0009] Wherein, the iterative denoising is to perform cyclic denoising on the decoded data by using the loop filter, and use the output data of the previous cycle as one of the input data of the current cycle in each cycle.

[0010] Optionally, the performing iterative denoising on the decoded data by using the loop filter includes:

[0011] Obtaining input data required for the current iteration; the input data required for the current iteration includes the output data of the previous iteration and the iteration order corresponding to the current iteration in the iterative denoising process, and the input data required for the first iteration includes the decoded data;

[0012] Send the input data required for the current iteration to the loop filter for denoising to obtain the denoising result of the current iteration output by the loop filter; the output data of each iteration is the denoising result output by the loop filter for the corresponding input data in each iteration.

[0013] A data processing method includes:

[0014] Obtain the encoded data obtained by compressing and encoding the video frames of the video to be processed; the image quality corresponding to the encoded data of the video frame is lower than the original image quality of the video frame.

[0015] Determine the target number of iterations required for denoising the decoded data corresponding to the encoded data; the target number of iterations is: when using a pre-constructed loop filter to perform iterative denoising on the decoded data, the number of iterations corresponding to the image quality of the denoising result satisfying the image quality condition.

[0016] Generate and output the compressed result data of the video to be processed.

[0017] Wherein, the compressed result data includes: the encoded data of each video frame of the video to be processed and the target number of iterations corresponding to each video frame respectively; the iterative denoising includes using the loop filter to perform cyclic denoising on the decoded data, and using the output data of the previous cycle as one of the input data of the current cycle in each cycle.

[0018] Optionally, determining the target number of iterations required for denoising the decoded data corresponding to the encoded data includes:

[0019] Determine the target distortion degree of the image quality corresponding to the encoded data relative to the image quality corresponding to the video frame.

[0020] According to the mapping relationship between different distortion degrees and different numbers of iterations formulated in advance, determine the number of iterations corresponding to the target distortion degree to obtain the target number of iterations required for denoising the decoded data corresponding to the encoded data.

[0021] Optionally, determining the target number of iterations required for denoising the decoded data corresponding to the encoded data includes:

[0022] Decode the encoded data to obtain the decoded data corresponding to the encoded data.

[0023] Use the loop filter to perform iterative denoising on the decoded data, and determine the image quality corresponding to the denoising result output by the loop filter in each iteration.

[0024] Determine the iteration order corresponding to the target iteration for which the image quality corresponding to the denoising result meets the image quality condition during the iterative denoising process, and determine the target number of iterations according to the iteration order.

[0025] Optionally, performing iterative denoising on the decoded data using the loop filter includes:

[0026] Obtain the input data required for the current iteration; the input data required for the current iteration includes the output data of the previous iteration and the iteration order corresponding to the current iteration during the iterative denoising process, and the input data required for the first iteration includes the decoded data;

[0027] Send the input data required for the current iteration to the loop filter for denoising to obtain the denoising result of the current iteration output by the loop filter; the output data of each iteration is the denoising result output by the loop filter for the corresponding input data in each iteration.

[0028] Optionally, determining the iteration order corresponding to the target iteration for which the image quality corresponding to the denoising result meets the image quality condition during the iterative denoising process includes:

[0029] Determine the image quality index values of the image qualities corresponding to the denoising results output by each iteration;

[0030] Determine the target iteration for which the image quality corresponding to the denoising result meets the image quality condition according to the image quality index values of the image qualities corresponding to the denoising results output by each iteration;

[0031] Determine the iteration order corresponding to the target iteration during the iterative denoising process.

[0032] Optionally, determining the target iteration for which the image quality corresponding to the denoising result meets the image quality condition according to the image quality index values of the image qualities corresponding to the denoising results output by each iteration includes:

[0033] Determine the iteration for which the image quality characterized by the image quality index value of the denoising result output in each iteration meets the image quality degradation condition compared to the image quality characterized by the image quality index value of the denoising result output in the corresponding previous iteration;

[0034] Determine the previous iteration corresponding to the iteration that meets the image quality degradation condition to obtain the target iteration.

[0035] Optionally, the loop filter is a neural network model obtained by training the preset neural network through iterative denoising of image samples.

[0036] A data processing device includes:

[0037] A first acquisition module, configured to obtain the compressed result data of the video to be processed; the compressed result data includes the encoded data of each video frame of the video to be processed and the target iteration times respectively corresponding to each video frame; the target iteration times are: when iteratively denoising the decoded data corresponding to the encoded data by using a pre-constructed loop filter, the iteration times that can make the image quality corresponding to the denoising result meet the image quality condition;

[0038] A decoding module, configured to decode the encoded data of the video frames in the compressed result data to obtain the corresponding decoded data;

[0039] A denoising module, configured to perform iterative denoising on the decoded data by using the loop filter to obtain a denoising result; wherein, the iteration times are the target iteration times;

[0040] An output module, configured to output the image corresponding to the denoising result obtained by performing iterative denoising;

[0041] Wherein, the iterative denoising is to perform cyclic denoising on the decoded data by using the loop filter, and use the output data of the previous cycle as one of the input data of the current cycle in each cycle.

[0042] A data processing device, comprising:

[0043] A second acquisition module, configured to obtain the encoded data obtained by performing compression encoding on the video frames of the video to be processed; the image quality corresponding to the encoded data of the video frames is lower than the original image quality of the video frames;

[0044] A determination module, configured to determine the target iteration times required for denoising the decoded data corresponding to the encoded data; the target iteration times are: when iteratively denoising the decoded data by using a pre-constructed loop filter, the iteration times corresponding to the denoising result that can make the image quality meet the image quality condition;

[0045] A generation and output processing module, configured to generate and output the compressed result data of the video to be processed;

[0046] Wherein, the compressed result data includes: the encoded data of each video frame of the video to be processed and the target iteration times respectively corresponding to each video frame; the iterative denoising includes performing cyclic denoising on the decoded data by using the loop filter, and using the output data of the previous cycle as one of the input data of the current cycle in each cycle.

[0047] As can be seen from the above solution, the present application discloses a data processing method and apparatus. The data processing method includes: obtaining compressed result data of a video to be processed; the compressed result data includes encoded data of each video frame of the video to be processed and target iteration times respectively corresponding to each video frame; the target iteration time is: when iteratively denoising decoded data corresponding to the encoded data by using a pre-constructed loop filter, the iteration time that can make the image quality corresponding to the denoising result meet the image quality condition; decoding the encoded data of the video frame in the compressed result data to obtain corresponding decoded data; using the loop filter to perform iterative denoising on the decoded data to obtain a denoising result; wherein, the iteration time is the target iteration time; outputting an image corresponding to the denoising result obtained by performing iterative denoising; wherein, the iterative denoising is to perform cyclic denoising on the decoded data by using the loop filter, and use the output data of the previous cycle as one of the input data of the current cycle each time. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0049] Figure 1 It is a schematic diagram of the operation mode of the loop filter provided by the present application;

[0050] Figure 2 It is a schematic diagram of the model network training of the loop filter provided by the present application;

[0051] Figure 3 It is a flowchart of a data processing method applied to the encoding end provided by the present application;

[0052] Figure 4 It is another flowchart of a data processing method applied to the encoding end provided by the present application;

[0053] Figure 5 It is a flowchart of determining the target iteration time required for iterative denoising provided by the present application;

[0054] Figure 6 It is a schematic diagram of the correlation between PSNR and the iteration order / times provided by the present application;

[0055] Figure 7 It is another flowchart of determining the target iteration time required for iterative denoising provided by the present application;

[0056] Figure 8 is a flow chart of a data processing method applied to a decoding end provided by the present application;

[0057] Figure 9 It is a structural diagram of a data processing device applied to an encoding end provided by the present application;

[0058] Figure 10 It is a structural diagram of a data processing device applied to a decoding end provided by the present application;

[0059] Figure 11 It is a structural diagram of the electronic device provided in this application. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0061] In order to improve the compression rate, lossy compression is usually used, so that the video can be highly compressed (such as 1000 to 10000 times). Lossy compression will discard high-frequency signals that are not sensitive to the human eye during the compression encoding process. Discarding high-frequency signals can reduce the amount of signals that need to be encoded and improve the compression rate, but it will also introduce distortion, resulting in a decrease in image quality. At present, there are some methods to reduce distortion and improve the image quality when decoding and outputting videos, such as using a neural network-like method to train filters and using filters to denoise the decoded data of video compression data, so as to achieve the effect of reducing distortion and improving image quality.

[0062] In the above solution, after the neural network training is completed, its network parameters are fixed immediately to obtain a fixed filter, and its corresponding denoising performance is also limited accordingly. However, the applicant found that in video compression, different video frames usually have different degrees of distortion. The differentiated noise caused by different degrees of distortion usually cannot achieve the required denoising performance with only a fixed filter. In other words, a fixed filter is difficult to apply to different degrees of distortion, and it is difficult to achieve the required denoising effect for different degrees of distortion.

[0063] Based on this, an embodiment of the present application discloses a data processing method and device, which can be applied to video compression / decompression scenarios, to solve the above-mentioned defects of known technologies in denoising different degrees of distortion caused by video compression, and to improve image quality while ensuring the required video compression rate.

[0064] The disclosed method includes a data processing method applied to the encoding end (i.e., the Encoder side, also known as the compression end) and a data processing method applied to the decoding end (i.e., the Decoder side, also known as the decompression end). Moreover, the disclosed data processing method applied to the encoding end / decoding end can be applied to, but is not limited to, electronic devices such as personal computers or servers.

[0065] The disclosed data processing method is premised on pre-constructing a filter applicable to denoising different degrees of distortion generated by video compression.

[0066] Among them, in this application, through neural network model training, a filter applicable to denoising different degrees of distortion generated by video compression is obtained. In order to be applicable to different degrees of distortion, that is, to achieve the required denoising effect for different degrees of distortion, in both the model training and inference stages, the model network is used iteratively. That is, in the model training / inference stage, when denoising, the same model network is used repeatedly, and in each loop, the output data of the previous loop is used as one of the input data of the current loop. By repeatedly introducing the output data of the previous loop into the input of the current loop during the loop process, iterative denoising is realized, and the distortion is gradually removed based on iterative denoising.

[0067] Among them, for images with different degrees of distortion, denoising can be achieved through different numbers of iterations to achieve the required image quality restoration effect.

[0068] In the embodiments of this application, the filter obtained through iterative denoising training is called a loop filter. More specifically, it can be called a Progressive in-loop filter (PIF).

[0069] Correspondingly, the loop filter is specifically a neural network model obtained by training a preset neural network through iterative denoising of image samples. The image samples can be video frame samples or image samples, without limitation.

[0070] The preset neural network can be, but is not limited to, UNet or ResNet.

[0071] See Figure 1 the schematic diagram of the operation mode of the loop filter shown, and Figure 2 the schematic diagram of the model network training shown, Figure 1 and Figure 2The DN (denoising network) therein represents a preset neural network, and the trained DN is the loop filter. t represents the iteration number corresponding to the current iteration. In each iteration, the corresponding iteration number can be used as one of the input data of the network to determine the distortion recovery under different distortion degrees.

[0072] Optionally, before training the DN, the golden data required for training can be prepared first. In machine learning, "Golden Data" usually refers to high-quality and highly accurate data sets, which play an important role in the training and evaluation of machine learning models. In this embodiment, the golden data includes the original image before compression, specifically including a series of uncompressed original video frames or original pictures, and the golden data is lossily compressed to obtain the picture data with noise (noise caused by distortion). The picture data with noise can be regarded as the sample data for training the DN.

[0073] On this basis, iterative denoising training is performed on the DN. In each iteration, the picture data with noise x_t is input into the DN, and at the same time, the iteration number t corresponding to the current iteration is input into the DN. Optionally, the MLP embedding method can be used to provide t to the DN. After the DN denoises the x_t, the corresponding denoising result x_(t + 1) is obtained, output, and the neural network DN is updated through the propagation of the reverse loss gradient, so that the network parameters of the DN are gradually optimized until the end condition is met, and the loop filter is obtained. The end condition can be, but is not limited to, set as the training duration reaching the set duration, or the number of training times of the DN reaching the set number, or the network loss being lower than the preset loss, etc.

[0074] In the first iteration, the picture data with noise x_t is the lossy picture data after compressing the golden data, and the corresponding iteration number t is specifically 0. In non-first iterations, x_t is the output data of the previous iteration, and the value of t increases sequentially according to the iteration order.

[0075] The loss function can include, but is not limited to, the PSNR (Peak signal-to-noise ratio) or MSE (mean-square error) between the golden data corresponding to the output data x_(t + 1) and the output data x_(t + 1). The larger the PSNR, the better the model performance, and the smaller the MSE, the better the model performance.

[0076] For example, Figure 2 In , the loss function specifically includes the MSE between the golden data corresponding to the output data x_(t+1) and the output data x_(t+1), expressed as MSE(golden data, x_(t+1)).

[0077] Based on the constructed loop filter, different numbers of iterations can be used for different degrees of distortion of each video frame after compression. The loop filter can be used to iteratively denoise each video frame with noise (noise caused by distortion) to achieve the technical effect of improving the image quality as much as possible and restoring the original image quality of each video frame.

[0078] Based on the constructed loop filter, see Figure 3 The data processing method applied to the encoding end disclosed in the present application at least includes the following processing steps:

[0079] Step 301: Obtain coded data obtained by compressing and encoding a video frame of a video to be processed; the image quality corresponding to the coded data of the video frame is lower than the original image quality of the video frame.

[0080] The compression coding of the video frame of the video to be processed is lossy compression.

[0081] Optionally, when compressing and encoding the video frame of the video to be processed, the difference between the current video frame of the video to be processed and the corresponding reference frame can be determined, and the obtained difference and identification information of the reference frame can be encoded to obtain the encoded data of the current video frame.

[0082] The identification information of the reference frame may include, but is not limited to, information such as the frame number or timestamp of the reference frame.

[0083] The reference frame corresponding to the current video frame is the corresponding frame of the video to be processed, or is empty. For example, for the first frame of the video to be processed, the corresponding reference frame is empty. When the reference frame corresponding to the current video frame is not empty, the reference frame is specifically a frame before the corresponding timestamp of the current video frame in the video to be processed, and preferably, the reference frame and the current video frame meet similar conditions to minimize the difference between the current video frame and the reference frame and reduce the amount of data encoding.

[0084] In the encoding process, the difference is converted into frequency domain data, and high-frequency signals in the frequency domain data that meet the high-frequency conditions are discarded. Discarding high-frequency signals can reduce the amount of signals that need to be encoded and improve the compression rate, but it will also introduce distortion, resulting in a decrease in image quality. The embodiment of the present application will subsequently use the constructed loop filter to iteratively denoise the video frame image data with noise caused by distortion to improve the image quality and restore the original image quality of the video frame as much as possible.

[0085] Step 302: Determine the target number of iterations required for denoising the decoded data corresponding to the encoded data; the target number of iterations is the number of iterations that can satisfy the image quality condition for the denoising result when iteratively denoising the decoded data using a pre-constructed loop filter.

[0086] Iteratively denoising the decoded data includes circularly denoising the decoded data using a loop filter, and using the output data of the previous loop as one of the input data for the current loop in each loop.

[0087] The image quality condition can be, but is not limited to, set such that the image quality index value of the denoising result is within a preset value range representing high image quality, or the image quality index value of the denoising result is the image quality index value representing the top_k image quality in each iteration. Where k≥1 and k is an integer.

[0088] Step 303: Generate and output the compressed result data of the video to be processed; the compressed result data includes: the encoded data of each video frame of the video to be processed and the target number of iterations corresponding to each video frame.

[0089] Optionally, for each video frame, the corresponding target number of iterations can also be encoded. The target number of iterations included in the compressed result data of the video to be processed can correspondingly be the encoded data of the target number of iterations.

[0090] In an alternative embodiment, the target number of iterations of the video frame can be independently encoded, and the encoded data of the video frame and the encoded data of the target number of iterations corresponding to the video frame can be encapsulated together, for example, encapsulated as a corresponding video stream (bitsream).

[0091] However, it is not limited to this. In other embodiments, the target number of iterations corresponding to the video frame can also be concatenated with the data to be encoded of the video frame first. For example, the target number of iterations corresponding to the video frame is added to the head or tail of the data to be encoded of the video frame, etc., and the concatenated result is encoded as a whole.

[0092] The data to be encoded of the video frame can include, but is not limited to, the difference between the video frame and the corresponding reference frame, and identification information such as the frame number of the reference frame.

[0093] In summary, for the data processing method applied to the encoding end provided in the embodiments of the present application, after obtaining the encoded data obtained by compressing and encoding the video frames of the video to be processed, by determining the target number of iterations required for denoising the decoded data corresponding to the encoded data of the video frames, and taking the target number of iterations corresponding to each video frame as a part of the compressed result data of the video to be processed, generating and outputting the compressed result data of the video to be processed, it is convenient to perform iterative denoising on the noisy picture data caused by distortion of each video frame according to the target number of iterations specified for each video frame in the compressed result data when decoding and outputting the compressed result data of the video to be processed later, and the picture quality corresponding to the denoising result can meet the picture quality condition.

[0094] Different degrees of distortion are based on different target numbers of iterations to achieve the denoising effect required by the set picture quality condition. And compared with the data of the video frames, the amount of data of the target number of iterations is very small. Introducing the target number of iterations (encoded data of the target number of iterations) corresponding to the video frames into the compressed result data of the video to be processed will not increase too much data volume, and still can ensure obtaining the required compression ratio (or a compression ratio highly approximated to the required compression ratio). Therefore, the present application overcomes the problems existing in the known technology, can be applicable to different degrees of distortion, and can respectively achieve the required denoising effect for different degrees of distortion, improving the picture quality on the premise of ensuring the required compression ratio.

[0095] In an optional embodiment, referring to Figure 4 the flowchart of the data processing method shown, the data processing method applied to the encoding end disclosed in the present application may further include the following processing before step 101:

[0096] Step 401: Obtain the video to be processed, and perform compression encoding on the video frames of the video to be processed to obtain the encoded data of the video frames.

[0097] In this embodiment, specifically, the video to be processed is first obtained at the encoding end, and based on the lossy compression method, the video frames of the video to be processed are compressed and encoded. Exemplarily, the video to be processed is obtained by the encoder, and the difference between the video frames of the video to be processed and the corresponding reference frames is predicted by the prediction unit of the encoder, and then the obtained difference and information such as the frame numbers of the reference frames are encoded by the encoding unit of the encoder to obtain the encoded data of the video frames.

[0098] During the encoding process, the difference between the video frame and its reference frame is converted into frequency domain data, and the high-frequency signals that meet the high-frequency condition in the frequency domain data are discarded.

[0099] In practical applications, not limited to the above embodiments, it is also possible to directly obtain the encoded data obtained by compressing and encoding the video frames of the video to be processed at the encoding end. For example, directly obtain the video frame encoding data of the video to be processed transmitted remotely. There is no limitation on this, and it can be determined according to the actual application requirements.

[0100] Subsequently, based on the processing procedures of steps 301-303, the final compressed result data of the video to be processed can be further generated and output.

[0101] Since the target iteration times corresponding to each video frame are introduced into the compressed result data generated and output for the video to be processed, when decoding and outputting the compressed result data of the video to be processed, it can guide the decoding end to perform iterative denoising on the noisy picture data caused by distortion of each video frame according to the target iteration times specified for each video frame in the compressed result data, and make the image quality corresponding to the denoising result meet the image quality conditions. Accordingly, the technical effect of improving the image quality can be achieved on the premise of ensuring the required compression ratio.

[0102] In an optional embodiment, referring to Figure 5 Step 302 in the data processing method applied to the encoding end disclosed in the present application can be implemented as the following processing procedure:

[0103] Step 501: Decode the encoding data of the video frames of the video to be processed to obtain the decoded data corresponding to the encoding data.

[0104] In this embodiment, specifically at the encoding end, the encoding data of the video frames of the video to be processed is decoded. Exemplarily, a decoding unit can be added to the encoder, and the added decoding unit is used to decode the encoding data of the video frames of the video to be processed. Through the decoding process, information such as the difference between the video frame of the video to be processed and its reference frame and the reference frame number is parsed out, and according to the difference and the picture data of the reference frame indicated by the reference frame number, the noisy (noise caused by distortion) picture data of the video frame is assembled, that is, the decoded data corresponding to the encoding data of the video frame.

[0105] Step 502: Use a loop filter to perform iterative denoising on the decoded data, and determine the image quality corresponding to the denoising result output by the loop filter each time.

[0106] After obtaining the decoded data corresponding to the encoding data of the video frame, the loop filter is further used to perform iterative denoising on the decoded data. Optionally, the loop filter can be set in the decoding unit of the encoder. After the decoding unit of the encoder decodes the encoding data of the video frames of the video to be processed to obtain the decoded data of the video frame, that is, the noisy picture data of the video frame, it is sent to the loop filter for iterative denoising.

[0107] In each iteration of iterative denoising, the input data required for the current iteration can be obtained first. The input data required for the current iteration includes the output data of the previous iteration and the iteration order corresponding to the current iteration in the iterative denoising process. The input data required for the first iteration includes the decoded data. Then, the input data required for the current iteration is sent to the loop filter for denoising to obtain the denoising result of the current iteration output by the loop filter. The output data of each iteration is the denoising result output by the loop filter for the corresponding input data in each iteration.

[0108] For example, referring to Figure 1 , for the current iteration, its input data x_t and the corresponding iteration number t can be obtained, and x_t and t are sent to the loop filter (DN that has completed training) for denoising to obtain x(t + 1) output by the loop filter. Subsequently, x(t + 1) and (t + 1) will be used as the input data for the next iteration and sent to the loop filter for denoising. In the first iteration, its input data includes the decoded data obtained by decoding the encoded data of the video frame. Optionally, it also includes the corresponding iteration order 0. In subsequent iterations, the iteration order increases sequentially.

[0109] In the iterative denoising process, this embodiment also determines the image quality corresponding to the denoising result output by the loop filter in each iteration. Specifically, the image quality index value corresponding to the denoising result can be determined, and the image quality corresponding to the denoising result is characterized based on the image quality index value. The image quality index value can be, but is not limited to, PSNR or MSE. The larger the PSNR, the higher the image quality; the smaller the MSE, the higher the image quality.

[0110] Step 503: Determine the iteration order corresponding to the target iteration in the iterative denoising process where the image quality corresponding to the denoising result meets the image quality condition, and determine the target iteration number according to the iteration order.

[0111] As described above, the image quality condition can be, but is not limited to, set as the image quality index value of the image quality corresponding to the denoising result being within the preset value range representing high image quality, or the image quality index value of the image quality corresponding to the denoising result being the image quality index value representing the top_k image quality in each iteration. Wherein, k ≥ 1 and k is an integer.

[0112] When determining the target iteration number, specifically, the image quality index values corresponding to the image qualities of the denoising results output in each iteration can be determined. According to the image quality index values corresponding to the image qualities of the denoising results output in each iteration, the target iteration where the image quality corresponding to the denoising result meets the image quality condition is determined, and the iteration order corresponding to the target iteration in the iterative denoising process is determined. Based on the iteration order corresponding to the target iteration in the iterative denoising process, the target iteration number required for denoising the decoded data corresponding to the encoded data of the video frame is determined.

[0113] Optionally, the iteration order can start from 0 and increase sequentially. For a certain iteration with the iteration order of x, during the overall iterative denoising process of the corresponding video frame, the corresponding number of iterations is x, where x is an integer not less than 0.

[0114] The following takes the image quality condition set as "the image quality index value corresponding to the denoising result is the image quality index value representing the top_1 image quality in each iteration" as an example to provide an exemplary implementation process for determining the target number of iterations.

[0115] The applicant found that during the iterative process, there is a correlation between the image quality index value corresponding to the denoising result and the iteration order / number of iterations. Taking PSNR as an example, the relationship between it and the iteration order / number of iterations is as Figure 6 shown. Initially, as the iteration order / number of iterations increases, the PSNR value continuously increases. When the iteration order / number of iterations increases to a certain value ( Figure 6 n in it), the PSNR value reaches the highest point, and then the PSNR value continuously decreases as the iteration order / number of iterations increases.

[0116] Based on this, specifically, it is possible to determine the iteration in which the image quality represented by the image quality index value corresponding to the denoising result output in each iteration, compared to the image quality represented by the image quality index value corresponding to the denoising result output in the previous corresponding iteration, satisfies the image quality degradation condition; and determine the previous iteration corresponding to the iteration that satisfies the image quality degradation condition to obtain the target iteration, and then based on the iteration order corresponding to the target iteration (such as Figure 6 n in it), determine the target number of iterations required for denoising the decoded data corresponding to the encoded data of the video frame.

[0117] Optionally, the image quality degradation condition can be set as follows: during the iterative process, as the iteration order / number of iterations increases, the image quality corresponding to the current iteration changes from higher than the image quality corresponding to its previous iteration to lower than the image quality corresponding to its previous iteration, that is, the phenomenon of image quality degradation starts to occur (such as the corresponding PSNR value starts to decrease). Based on the set image quality degradation condition, the previous iteration corresponding to the current iteration when the image quality degradation phenomenon is first discovered can be determined as the target iteration, and then the iteration order corresponding to the target iteration in the iterative denoising process is used as the target number of iterations required for denoising the decoded data corresponding to the encoded data of the corresponding video frame (such as Figure 6 n in it).

[0118] In practical applications, there may be fluctuations in the change law of the image quality corresponding to different-order iterations during the iterative process. Taking the image quality index value as PSNR as an example, PSNR may not completely follow Figure 6Based on the shown variation pattern, in other embodiments, the image quality degradation condition can be set such that during the iteration process, as the iteration order / number increases, the image quality corresponding to the current iteration changes from being higher than the image quality corresponding to the previous iteration to being lower than the image quality corresponding to the previous iteration, and the image quality continues to degrade during a preset number of iterations after the current iteration. By setting this image quality degradation condition, the determined target iteration number is prevented from being affected by the above-mentioned fluctuation phenomenon, ensuring the accuracy of the determined target iteration number.

[0119] In an alternative embodiment, referring to Figure 7 , step 302 in the data processing method applied to the encoding end disclosed in the present application can also be implemented as the following processing procedure:

[0120] Step 701: Determine the target distortion degree of the image quality corresponding to the encoded data of the video frames of the video to be processed relative to the image quality corresponding to the video frames.

[0121] In this step, the image quality corresponding to the video frame is the original image quality corresponding to the video frame before compression encoding.

[0122] The target distortion degree can be characterized based on the image quality index value of the decoded data corresponding to the encoded data of the video frames of the video to be processed. The higher the image quality reflected by the image quality index value, the lower the characterized distortion degree.

[0123] Step 702: According to the pre-established mapping relationship between different distortion degrees and different iteration numbers, determine the iteration number corresponding to the target distortion degree, and obtain the target iteration number required for denoising the decoded data corresponding to the encoded data.

[0124] In this embodiment, a mapping relationship between different distortion degrees and different iteration numbers is pre-established based on the iterative denoising process of the loop filter. For each distortion degree, this mapping relationship specifies the iteration number that can make the image quality corresponding to the denoising result meet the image quality condition when using the loop filter to perform iterative denoising on the decoded data of the video frame encoded data.

[0125] The mapping relationship between different distortion degrees and different iteration numbers can specifically be a corresponding relationship (one-to-one, one-to-many, or many-to-one) between distortion degree values and iteration number values, or it can also be a mapping relationship between a distortion degree range / interval and an iteration number value, a mapping relationship between a distortion degree value and an iteration number range / interval, etc. There is no limitation on this, and it can be determined according to the actual situation.

[0126] After determining the target distortion degree, correspondingly, based on the pre-established mapping relationship between different distortion degrees and different numbers of iterations, the number of iterations corresponding to the target distortion degree can be determined, and this number of iterations is used as the target number of iterations required for denoising the decoded data corresponding to the encoded data of the video frame. When the mapping relationship is the corresponding relationship between the distortion degree value and the range / interval of the number of iterations, an arbitrary number of iterations can be selected from the range / interval of the number of iterations corresponding to the target distortion degree as the target number of iterations, or a number of iterations can also be selected from this range / interval of the number of iterations corresponding to the target distortion degree based on a certain strategy (such as selecting the median or the minimum value, etc.) as the target number of iterations.

[0127] Video compression using different compression methods usually results in different degrees of distortion of the video frame. That is to say, different compression methods usually correspond to different distortion degrees. Based on this, in practical applications, optionally, the mapping relationship between different compression methods and different numbers of iterations can be pre-established based on the iterative denoising process of the loop filter to represent the mapping relationship between different distortion degrees and different numbers of iterations. Similarly, the mapping relationship between different compression methods and different numbers of iterations established can specifically be the corresponding relationship between the compression method and the number of iteration values (one-to-one, one-to-many, or many-to-one), or it can also be the mapping relationship between the compression method and the range / interval of the number of iterations, etc. There is no limitation on this and it can be determined according to the actual situation.

[0128] In this embodiment, when determining the target distortion degree of the image quality corresponding to the encoded data of the video frame of the video to be processed relative to the image quality corresponding to the video frame, specifically, the target compression method used for compressing the video to be processed can be determined, the target distortion degree is characterized based on the target compression method, and based on the pre-established mapping relationship between different compression methods and different numbers of iterations, the number of iterations corresponding to the target compression method is determined as the target number of iterations. When the pre-established mapping relationship is the corresponding relationship between the compression method and the range / interval of the number of iterations, an arbitrary number of iterations can be selected from the range / interval of the number of iterations corresponding to the target compression method as the target number of iterations, or a number of iterations can also be selected from this range / interval of the number of iterations corresponding to the target compression method based on a certain strategy (such as selecting the median or the minimum value, etc.) as the target number of iterations.

[0129] Subsequently, by introducing the target number of iterations corresponding to each video frame into the compressed result data of the video to be processed, the decoding end can be guided to perform iterative denoising on the noisy picture data caused by distortion of each video frame according to the target number of iterations specified for each video frame in the compressed result data, so that the image quality corresponding to the denoising result of each video frame meets the image quality condition, achieving the technical effect of improving the image quality while ensuring the required compression ratio.

[0130] See Figure 8 The data processing method applied to the decoding end shown below. The data processing method applied to the decoding end disclosed in this application includes the following processing steps:

[0131] Step 801: Obtain the compressed result data of the video to be processed; the compressed result data includes the encoded data of each video frame of the video to be processed and the target iteration times corresponding to each video frame respectively.

[0132] The target iteration times are: when iteratively denoising the decoded data corresponding to the encoded data by using a pre-constructed loop filter, the iteration times that can make the image quality corresponding to the denoising result meet the image quality condition.

[0133] Specifically, the decoder can obtain the compressed result data of the video to be processed.

[0134] Step 802: Decode the encoded data of the video frames in the compressed result data to obtain the corresponding decoded data.

[0135] In the compressed result data, the encoded data of each video frame includes the differential encoding result between the video frame and the corresponding reference frame, and the encoded result of the identification information (such as the reference frame number) of the corresponding reference frame.

[0136] After obtaining the compressed result data of the video to be processed, the decoding unit of the decoder can be used to decode the encoded data of the video frames in the compressed result data. By decoding, information such as the difference between the video frame of the video to be processed and its corresponding reference frame and the reference frame number can be parsed out, and according to the difference and the picture data of the reference frame indicated by the reference frame number, the noisy (noise caused by distortion) picture data of the video frame can be assembled, that is, the decoded data corresponding to the encoded data of the video frame.

[0137] The target iteration times corresponding to each video frame can also be in the form of encoded data in the compressed result data. Correspondingly, the target iteration times corresponding to each video frame can be obtained through the decoding function of the decoding unit in the decoder.

[0138] Step 803: Use the loop filter to perform iterative denoising on the decoded data to obtain a denoising result; where the iteration times are the target iteration times.

[0139] A loop filter is set in the decoder.

[0140] After obtaining the decoded data corresponding to the encoded data of the video frame, further use the loop filter set in the decoder to perform iterative denoising on the decoded data, and the iteration times are the target iteration times corresponding to the video frame.

[0141] Optionally, the loop filter may be set in the decoding unit of the decoder. After the decoding unit of the decoder decodes the encoded data of the video frame of the video to be processed to obtain the decoded data of the video frame, that is, the noisy picture data of the video frame, it is sent to the loop filter for iterative denoising.

[0142] The iterative denoising includes using the loop filter to perform cyclic denoising on the decoded data, and using the output data of the previous cycle as one of the input data of the current cycle in each cycle.

[0143] Among them, the number of denoising times for performing iterative denoising on the decoded data using the loop filter is the target number of iterations. By performing denoising on the decoded data for the target number of iterations, the image quality corresponding to the obtained denoising result meets the image quality condition. The denoising process of the iterative denoising can be further implemented as the following steps a - step b:

[0144] Step a: Obtain the input data required for the current iteration; the input data required for the current iteration includes the output data of the previous iteration and the iteration order corresponding to the current iteration in the iterative denoising process. The input data required for the first iteration includes the decoded data.

[0145] Step b: Send the input data required for the current iteration to the loop filter for denoising to obtain the denoising result of the current iteration output by the loop filter; the output data of each iteration is the denoising result output by the loop filter for the corresponding input data in each iteration.

[0146] For example, with reference to Figure 1 , for the current iteration, its input data x_t and the corresponding iteration number t can be obtained, and x_t and t are sent to the loop filter (the trained DN) for denoising processing to obtain x_(t + 1) output by the loop filter. Subsequently, x_(t + 1) and (t + 1) will be used as the input data for the next iteration and sent to the loop filter for denoising. In the first iteration, its input data includes the decoded data obtained by decoding the encoded data of the video frame to be processed. Optionally, it also includes the corresponding iteration order 0. In subsequent iterations, the iteration order increases sequentially.

[0147] Step 804: Output the picture corresponding to the denoising result obtained by performing iterative denoising.

[0148] Based on step 803, finally, the picture corresponding to the denoising result can be output to meet the user's video picture viewing requirements.

[0149] In this embodiment, since the decoding end is guided by the target iteration times respectively specified for each video frame in the compressed result data of the video, and iteratively denoises the noisy picture data caused by distortion of each video frame, the picture quality corresponding to the denoising result of each video frame can meet the picture quality condition, thereby achieving the technical effect of improving the picture quality on the premise of ensuring the compression ratio required by the encoding end.

[0150] Corresponding to the data processing method applied to the encoding end above, an embodiment of the present application further provides a data processing device applied to the encoding end, and its composition structure is as Figure 9 shown, including:

[0151] A second acquisition module 901, configured to acquire the encoded data obtained by performing compression encoding on the video frames of the video to be processed; the picture quality corresponding to the encoded data of the video frame is lower than the original picture quality of the video frame;

[0152] A determination module 902, configured to determine the target iteration times required for denoising the decoded data corresponding to the encoded data; the target iteration times are: when iteratively denoising the decoded data by using a pre-constructed loop filter, the iteration times corresponding to the picture quality condition that can be satisfied by the denoising result;

[0153] A generation and output processing module 903, configured to generate and output the compressed result data of the video to be processed.

[0154] Wherein, the compressed result data includes: the encoded data of each video frame of the video to be processed and the target iteration times respectively corresponding to each video frame; the iterative denoising includes using the loop filter to perform cyclic denoising on the decoded data, and using the output data of the previous cycle as one of the input data of the current cycle in each cycle.

[0155] In an optional implementation manner, the determination module 902 is specifically configured to:

[0156] Determine the target distortion degree of the picture quality corresponding to the encoded data relative to the picture quality corresponding to the video frame;

[0157] According to the mapping relationship between different distortion degrees and different iteration times formulated in advance, determine the iteration times corresponding to the target distortion degree, and obtain the target iteration times required for denoising the decoded data corresponding to the encoded data.

[0158] In an optional implementation manner, the determination module 902 is specifically configured to:

[0159] Decode the encoded data to obtain the decoded data corresponding to the encoded data;

[0160] Iteratively denoise the decoded data using the loop filter, and determine the image quality corresponding to the denoising result output by the loop filter for each iteration.

[0161] Determine the iteration order corresponding to the target iteration in the iterative denoising process where the image quality corresponding to the denoising result meets the image quality condition, and determine the target number of iterations according to the iteration order.

[0162] In an alternative embodiment, the determination module 902, when iteratively denoising the decoded data using the loop filter, is specifically configured to:

[0163] Obtain the input data required for the current iteration; the input data required for the current iteration includes the output data of the previous iteration and the iteration order corresponding to the current iteration in the iterative denoising process, and the input data required for the first iteration includes the decoded data.

[0164] Send the input data required for the current iteration to the loop filter for denoising to obtain the denoising result of the current iteration output by the loop filter; the output data of each iteration is the denoising result output by the loop filter for the corresponding input data in each iteration.

[0165] In an alternative embodiment, the determination module 902, when determining the iteration order corresponding to the target iteration in the iterative denoising process where the image quality corresponding to the denoising result meets the image quality condition, is specifically configured to:

[0166] Determine the image quality index values corresponding to the image quality of the denoising results output for each iteration.

[0167] Determine the target iteration where the image quality corresponding to the denoising result meets the image quality condition according to the image quality index values corresponding to the image quality of the denoising results output for each iteration.

[0168] Determine the iteration order corresponding to the target iteration in the iterative denoising process.

[0169] In an alternative embodiment, the determination module 902, when determining the target iteration where the image quality corresponding to the denoising result meets the image quality condition according to the image quality index values corresponding to the image quality of the denoising results output for each iteration, is specifically configured to:

[0170] Determine the iterations where the image quality characterized by the image quality index value corresponding to the denoising result output in each iteration meets the image quality degradation condition compared to the image quality characterized by the image quality index value corresponding to the denoising result output in the corresponding previous iteration.

[0171] Determine the previous iteration corresponding to the iteration that meets the image quality degradation condition to obtain the target iteration.

[0172] In an alternative embodiment, the loop filter is a neural network model obtained by training a preset neural network through iterative denoising of image samples using the preset neural network.

[0173] Corresponding to the data processing method applied to the decoding end described above, an embodiment of the present application further provides a data processing device applied to the decoding end, and its composition structure is as Figure 10 shown, including:

[0174] A first acquisition module 1001, configured to obtain compressed result data of a video to be processed; the compressed result data includes encoded data of each video frame of the video to be processed and a target iteration number corresponding to each video frame; the target iteration number is: when performing iterative denoising on the decoded data corresponding to the encoded data by using a pre-constructed loop filter, the iteration number that can make the image quality corresponding to the denoising result meet the image quality condition;

[0175] A decoding module 1002, configured to decode the encoded data of the video frame in the compressed result data to obtain corresponding decoded data;

[0176] A denoising module 1003, configured to perform iterative denoising on the decoded data by using the loop filter to obtain a denoising result; wherein, the iteration number is the target iteration number;

[0177] An output module 1004, configured to output a picture corresponding to the denoising result obtained by performing iterative denoising.

[0178] Wherein, the iterative denoising is to perform cyclic denoising on the decoded data by using the loop filter, and use the output data of the previous cycle as one of the input data of the current cycle in each cycle.

[0179] An embodiment of the present application also discloses an electronic device, and the composition structure of the electronic device is as Figure 11 shown, at least including:

[0180] A memory 10, configured to store a computer instruction set;

[0181] The computer instruction set can be implemented in the form of a computer program.

[0182] A processor 20, configured to implement the data processing method applied to the encoding end or the data processing method applied to the decoding end disclosed in any of the above method embodiments by executing the computer instruction set.

[0183] The processor 20 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a neural network processor (NPU), a deep learning processor (DPU), or other programmable logic devices, etc.

[0184] The electronic device is equipped with a display device and / or has a display interface and can be externally connected to a display device.

[0185] Optionally, the electronic device further includes a camera component and / or is connected to an external camera component.

[0186] In addition, the electronic device may further include components such as a communication interface and a communication bus. The memory, the processor, and the communication interface complete communication with each other through the communication bus.

[0187] The communication interface is used for communication between the electronic device and other devices. The communication bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc.

[0188] It should be noted that each embodiment in this specification is described in a progressive manner. The key point of each embodiment is the difference from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0189] For the convenience of description, when describing the above system or device, it is divided into various modules or units according to functions for description. Of course, when implementing the present application, the functions of each unit can be realized in the same or multiple software and / or hardware.

[0190] From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments of the present application.

[0191] Finally, it should also be noted that in this text, relational terms such as first, second, third, and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

[0192] The above are only the preferred embodiments of the present application. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A data processing method, comprising: Obtaining compressed result data of a video to be processed; the compressed result data includes encoded data of each video frame of the video to be processed and target iteration times respectively corresponding to each video frame; The target iteration time is: when iteratively denoising the decoded data corresponding to the encoded data by using a pre-constructed loop filter, the iteration time that can make the image quality corresponding to the denoising result meet the image quality condition; Decoding the encoded data of the video frame in the compressed result data to obtain corresponding decoded data; Performing iterative denoising on the decoded data by using the loop filter to obtain a denoising result; wherein, the iteration time is the target iteration time; Outputting a picture corresponding to the denoising result obtained by performing iterative denoising; Wherein, the iterative denoising is to perform cyclic denoising on the decoded data by using the loop filter, and use the output data of the previous cycle as one of the input data of the current cycle in each cycle.

2. The data processing method according to claim 1, wherein the performing iterative denoising on the decoded data by using the loop filter comprises: Obtaining input data required for the current iteration; The input data required for the current iteration includes the output data of the previous iteration and the iteration order corresponding to the current iteration in the iterative denoising process, and the input data required for the first iteration includes the decoded data; Sending the input data required for the current iteration to the loop filter for denoising to obtain the denoising result of the current iteration output by the loop filter; The output data of each iteration is the denoising result output by the loop filter for the corresponding input data in each iteration.

3. A data processing method, comprising: Obtaining encoded data obtained by performing compression encoding on video frames of a video to be processed; The image quality corresponding to the encoded data of the video frame is lower than the original image quality of the video frame; Determining the target iteration time required for denoising the decoded data corresponding to the encoded data; the target iteration time is: when iteratively denoising the decoded data by using a pre-constructed loop filter, the iteration time corresponding to the image quality that can make the denoising result meet the image quality condition; Generating and outputting the compressed result data of the video to be processed; Wherein, the compressed result data includes: the encoded data of each video frame of the video to be processed and the target iteration times respectively corresponding to each video frame; the iterative denoising includes performing cyclic denoising on the decoded data by using the loop filter, and using the output data of the previous cycle as one of the input data of the current cycle in each cycle.

4. The data processing method according to claim 3, wherein the determining the target iteration time required for denoising the decoded data corresponding to the encoded data comprises: Determining the target distortion degree of the image quality corresponding to the encoded data relative to the image quality corresponding to the video frame; According to the pre-established mapping relationship between different distortion degrees and different iteration times, determining the iteration time corresponding to the target distortion degree to obtain the target iteration time required for denoising the decoded data corresponding to the encoded data.

5. The data processing method according to claim 3, wherein the determining the target number of iterations required for denoising the decoded data corresponding to the encoded data includes: Decoding the encoded data to obtain the decoded data corresponding to the encoded data; Iteratively denoising the decoded data by using the loop filter, and determining the image quality corresponding to the denoising result output by the loop filter each time; Determining the iteration order corresponding to the target iteration in the iterative denoising process where the image quality corresponding to the denoising result meets the image quality condition, and determining the target number of iterations according to the iteration order.

6. The data processing method according to claim 5, wherein the determining the iteration order corresponding to the target iteration in the iterative denoising process where the image quality corresponding to the denoising result meets the image quality condition includes: Determining the image quality index values of the image quality corresponding to the denoising results output respectively in each iteration; Determining the target iteration where the image quality corresponding to the denoising result meets the image quality condition according to the image quality index values of the image quality corresponding to the denoising results output respectively in each iteration; Determining the iteration order corresponding to the target iteration in the iterative denoising process.

7. The data processing method according to claim 6, wherein the determining the target iteration where the image quality corresponding to the denoising result meets the image quality condition according to the image quality index values of the image quality corresponding to the denoising results output respectively in each iteration includes: Determining the iteration where the image quality characterized by the image quality index value of the image quality corresponding to the denoising result output in each iteration meets the image quality degradation condition compared with the image quality characterized by the image quality index value of the image quality corresponding to the denoising result output in the previous corresponding iteration; Determining the previous iteration corresponding to the iteration that meets the image quality degradation condition to obtain the target iteration.

8. The data processing method according to claim 1, wherein the loop filter is a neural network model obtained by training the preset neural network through iterative denoising of image samples.

9. A data processing device, comprising: A first acquisition module, configured to obtain the compressed result data of the video to be processed; the compressed result data includes the encoded data of each video frame of the video to be processed and the target number of iterations corresponding to each video frame respectively; the target number of iterations is: the number of iterations that can make the image quality corresponding to the denoising result meet the image quality condition when iteratively denoising the decoded data corresponding to the encoded data by using a pre-constructed loop filter; A decoding module, configured to decode the encoded data of the video frame in the compressed result data to obtain the corresponding decoded data; A denoising module, configured to perform iterative denoising on the decoded data by using the loop filter to obtain a denoising result; wherein, the number of iterations is the target number of iterations; An output module, configured to output the picture corresponding to the denoising result obtained by performing iterative denoising; Wherein, the iterative denoising is to perform cyclic denoising on the decoded data by using the loop filter, and use the output data of the previous cycle as one of the input data of the current cycle in each cycle.

10. A data processing device, comprising: A second acquisition module, configured to acquire encoded data obtained by performing compression encoding on video frames of a video to be processed; The image quality corresponding to the encoded data of the video frames is lower than the original image quality of the video frames; A determination module, configured to determine a target number of iterations required for denoising the decoded data corresponding to the encoded data; the target number of iterations is: when iteratively denoising the decoded data by using a pre-constructed loop filter, the number of iterations corresponding to the image quality of the denoising result satisfying the image quality condition; A generation and output processing module, configured to generate and output compressed result data of the video to be processed; Wherein, the compressed result data includes: the encoded data of each video frame of the video to be processed and the target number of iterations corresponding to each video frame respectively; the iterative denoising includes performing cyclic denoising on the decoded data by using the loop filter, and using the output data of the previous cycle as one of the input data of the current cycle in each cycle.

Images