Video image enhancement method, system and device and storage medium

Abstract

The invention discloses a video image enhancement method, system and device and a storage medium, and the key points of the technical scheme are that a video is segmented according to a pixel difference to obtain a plurality of target sub-videos; distributing each target sub-video to a corresponding node, and performing image enhancement processing on the received target sub-video by the node to obtain an enhanced target sub-video; and receiving the enhanced target sub-videos sent by each node, and splicing all the enhanced target sub-videos to obtain an image enhanced video. According to the method, the dependence of an algorithm on hardware is effectively reduced, and the processing speed of image enhancement is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of video image processing, and particularly relates to a video image enhancement method, system, device and storage medium. BACKGROUND

[0002] Image enhancement refers to a technology of improving image quality, highlighting useful information in an image and improving visual effects of the image through certain algorithms and processing means.

[0003] Existing image enhancement methods mainly depend on hardware performance, and have relatively high requirements for CPUs and graphics cards. If a CPU with low computing power is used, the operation speed is slow, and the processing speed of the image is also slow. SUMMARY

[0004] The application aims to provide a video image enhancement method, system, device and storage medium, which effectively reduces the dependence of algorithms on hardware and improves the processing speed of image enhancement.

[0005] The first aspect of the application provides a video image enhancement method, comprising:

[0006] dividing the video into a plurality of target sub-videos according to pixel differences;

[0007] allocating each target sub-video to a corresponding node, and performing image enhancement processing on the received target sub-video by the node to obtain an enhanced target sub-video;

[0008] receiving the enhanced target sub-video sent by each node, splicing all the enhanced target sub-videos to obtain an image enhanced video.

[0009] In some embodiments, the dividing the video into a plurality of target sub-videos according to pixel differences comprises:

[0010] determining a starting division point of the video, and determining a pixel difference calculation point according to the starting division point;

[0011] calculating the pixel difference between adjacent frames in the video from the pixel difference calculation point;

[0012] comparing each pixel difference with a preset threshold to obtain the correlation between adjacent frames corresponding to each pixel difference;

[0013] dividing the video according to the correlation between adjacent frames to obtain a plurality of target sub-videos.

[0014] In some embodiments, the determining a starting division point of the video and determining a pixel difference calculation point according to the starting division point comprises:

[0015] calculate an average pixel difference within a preset number of frames from a starting frame of the video;

[0016] determine whether the average pixel difference is greater than a threshold value, if yes, take a frame after a to-be-selected frame as the starting frame, the to-be-selected frame being a last frame within the preset number of frames, and perform the step of calculating the average pixel difference within the preset number of frames from the starting frame of the video, if no, take the starting frame as a starting split point, and obtain a pixel difference calculation point by adding the starting split point by the preset number of frames.

[0017] In some embodiments, the received target sub-video is subjected to image enhancement processing to obtain an enhanced target sub-video, including:

[0018] determine a key frame of the target sub-video;

[0019] perform time sequence arrangement on the frames of the target sub-video after the key frame is determined to obtain a to-be-processed sequence, and perform standardization processing on the to-be-processed sequence to obtain a standardized sequence;

[0020] input the standardized sequence into a trained image enhancement model to obtain a video enhancement frame.

[0021] In some embodiments, the image enhancement model includes:

[0022] a convolutional neural network for spatial feature extraction on the standardized sequence to obtain a feature sequence;

[0023] a long short-term memory network for analyzing the time sequence relationship between the key frame and the remaining frames in the feature sequence to obtain a time sequence analysis result;

[0024] a background frame determination module for determining a frame as a background frame if the frame is within a preset continuous number of frames in the feature sequence and the inter-frame pixel difference value is within a preset range;

[0025] an image denoising module for denoising processing on the feature sequence according to the time sequence analysis result to obtain a denoised feature sequence, selecting a target frame from the continuous background frames every preset number of frames under the condition of displacement correction on the continuous background frames, determining a similar frame corresponding to the target frame, and denoising processing on the target frame, and denoising processing on the similar frame corresponding to the target frame using the denoising parameters of the target frame;

[0026] an image super-resolution module for displacement correction on the denoised sequence feature according to the time sequence analysis result and a motion compensation algorithm to obtain an enhanced target sub-video, and displacement correction on the target frame under the condition of displacement correction on the continuous background frames, and displacement correction on the similar frame corresponding to the target frame using the displacement correction parameters of the target frame.

[0027] In some embodiments, after the image-enhanced video is obtained, the method comprises:

[0028] In a case where the frame rate of the image-enhanced video is less than a preset frame rate, according to the time sequence analysis result and the number of dynamic pictures, a dynamic compensation algorithm is used to insert a compensation frame, so as to obtain a frame-compensated image-enhanced video.

[0029] In some embodiments, the splicing of all the enhanced target sub-videos to obtain the image-enhanced video comprises:

[0030] The all enhanced target sub-videos are spliced according to time sequence, and in the splicing process, a preset number of frames are extended before and after a splicing point to obtain an extended frame segment corresponding to the splicing point;

[0031] The extended frame segment is smoothed to obtain the image-enhanced video.

[0032] The second aspect of the present application provides a video image enhancement system, comprising:

[0033] A video segmentation module 10 is configured to segment a video according to pixel difference to obtain a plurality of target sub-videos;

[0034] A video distribution module 20 is configured to distribute each target sub-video to a corresponding node, and the node is configured to perform image enhancement processing on the received target sub-video to obtain an enhanced target sub-video;

[0035] A receiving and splicing module 30 is configured to receive the enhanced target sub-video sent by each node, splice all the enhanced target sub-videos to obtain an image-enhanced video.

[0036] The third aspect of the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0037] The fourth aspect of the present application provides a computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the above method.

[0038] The technical solution provided by the present application has the following advantages and effects: the video is segmented according to pixel difference, and the segmented target sub-videos are concurrently distributed to corresponding nodes, which effectively reduces the dependence of the algorithm on hardware and improves the processing speed of image enhancement. By taking pixel difference as the basis for segmentation, the remaining frames in the target sub-video can be processed according to the key frames after the key frames are selected, and the response speed is fast. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1is a flowchart of a video image enhancement method provided by the present application;

[0040] Figure 2 is a structural block diagram of a video image enhancement system provided by the present application;

[0041] Figure 3 is an internal structure diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to facilitate the understanding of the present application, specific embodiments of the present application will be described in more detail below with reference to the accompanying drawings.

[0043] Unless specifically stated or defined otherwise, "first, second, …" used herein is merely used for differentiation of names, and does not represent a specific quantity or order.

[0044] Unless specifically stated or defined otherwise, the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0045] It should be noted that "fixed to" and "connected to" herein can be directly fixed or connected to an element, or indirectly fixed or connected to an element.

[0046] The video image enhancement method of the present embodiment is mainly used for image enhancement of monitoring video, and a multi-machine distributed cluster mode is used to construct a video image enhancement platform. The platform is composed of multiple high-performance servers, each server is equipped with a high-performance CPU and multiple high-graphics-memory GPUs such as NVIDIA A100 or H100 GPUs, which provide strong computing power support for image data processing. At the same time, the servers are interconnected through high-speed networks such as 100G Ethernet or InfiniBand network, which ensures that data can be quickly transmitted and shared, and meets the needs of large data interaction in distributed computing.

[0047] As shown in Figure 1 A video image enhancement method provided in the present embodiment includes the following steps S1-S3:

[0048] Step S1, the video is divided into multiple target sub-videos according to the pixel difference.

[0049] In actual application, a server is selected from the video image enhancement platform for video division to obtain multiple target sub-videos, so as to be subsequently distributed to the corresponding remaining servers for image enhancement processing.

[0050] Specifically, the video is divided into multiple target sub-videos according to the pixel difference, including:

[0051] determine a start split point of the video, and determine a pixel difference calculation point according to the start split point;

[0052] calculate pixel differences between adjacent frames in the video from the pixel difference calculation point;

[0053] compare each of the pixel differences with a preset threshold to obtain a correlation between adjacent frames corresponding to each of the pixel differences;

[0054] split the video according to the correlation between each of the adjacent frames to obtain a plurality of target sub-videos.

[0055] In actual application, for example, in a video, the fourth frame is taken as the pixel difference calculation point, the pixel value sum of all pixel points of the fourth frame is calculated, the pixel value sum of all pixel points of the third frame is calculated, the difference between the pixel value sum of the third frame and the fourth frame is calculated to obtain the pixel difference between the third frame and the fourth frame. Through the setting of the preset threshold, the correlation between the images of the frames can be determined, which provides a basis for video slicing.

[0056] In the case where the pixel difference exceeds the preset threshold, it indicates that the correlation between the images of adjacent frames is low, and in the case where the pixel difference does not exceed the preset threshold, it indicates that the correlation between the images of adjacent frames is high. The adjacent frames with low correlation are taken as the split points to obtain a plurality of selected sub-videos after splitting. It is judged whether the frame number of the selected sub-video is within a preset processing frame number range. If yes, the selected sub-video is taken as the target sub-video. If the frame number of the selected sub-video is lower than the preset processing frame number range, the pixel differences corresponding to the front split point and the rear split point of the selected sub-video are compared, and the selected sub-video is merged into the selected sub-video corresponding to the split point with smaller pixel difference to obtain the target sub-video. If the frame number of the selected sub-video is higher than the preset processing frame number range, the preset threshold is increased, and the selected sub-video is further split until the selected sub-video after splitting is within the preset processing frame number range. By taking the pixel difference between adjacent frames as the slicing basis, the key frame is selected in the subsequent process, so that the remaining frames in the target sub-video can be processed according to the key frame, and the response speed is fast.

[0057] Specifically, the determination of the start split point of the video and the determination of the pixel difference calculation point according to the start split point include:

[0058] calculate an average pixel difference within a preset frame number from the start frame of the video;

[0059] determine whether the average pixel difference is greater than a threshold. If yes, the next frame of the start frame is taken as the start frame, and the step of calculating the average pixel difference within the preset frame number from the start frame of the video is performed. If no, the start frame is taken as the start split point, and the pixel difference calculation point is obtained by adding the preset frame number to the start split point.

[0060] In practical applications, by calculating the average pixel difference within the preset frame number, it can be prevented that too many frames are discarded due to large pixel difference between the first few frames of the video. The preset frame number can be 3 frames, 4 frames, etc. If the preset frame number is 3 frames, the 3 frames are the 1st frame, the 2nd frame and the 3rd frame respectively. In this case, if the average pixel difference within the preset frame number is greater than the threshold value, 2 frames are taken as the starting frame, the 1st frame is discarded, the average pixel difference within the preset frame number is calculated from the 2nd frame, and the cycle is repeated in turn until the average pixel difference within the preset frame number is less than or equal to the threshold value. The starting frame at this time is taken as the starting split point. The pixel difference calculation point is obtained by adding the preset frame number to the starting split point. In this case, the starting frame is the 2nd frame, and the pixel difference calculation point is the 5th frame. The pixel difference between adjacent frames is calculated from the 5th frame, such as the pixel difference between the 5th frame and the 4th frame. For the calculation of the average pixel difference, within the preset frame number, the average pixel difference of all pixel points between adjacent frames can be calculated. From the pixel difference calculation point, the average pixel difference of part of the pixel points between adjacent frames can be calculated, such as the average pixel difference of the first 128 rows of pixel points, to reduce the amount of calculation and improve the operation efficiency.

[0061] In step S2, each target sub-video is assigned to a corresponding node, and the node performs image enhancement processing on the received target sub-video to obtain an enhanced target sub-video.

[0062] In practical applications, after the video is cut to obtain a plurality of target sub-videos, the plurality of target sub-videos are sequentially assigned to corresponding servers, i.e. corresponding nodes. The GPU of each server independently and in parallel processes the assigned target sub-video. By utilizing the parallel computing capability of multiple machines and multiple GPUs, the processing efficiency is greatly improved. During the processing, each server enhances image details and improves resolution by processing the target sub-video.

[0063] Specifically, the image enhancement processing on the received target sub-video to obtain an enhanced target sub-video includes:

[0064] determining a key frame of the target sub-video;

[0065] sequentially arranging the frames of the target sub-video after determining the key frame to obtain a to-be-processed sequence, and performing standardization processing on the to-be-processed sequence to obtain a standardized sequence;

[0066] inputting the standardized sequence into a trained image enhancement model to obtain a video enhancement frame.

[0067] In practical applications, a frame with high brightness or clear edge texture is selected as the key frame t, and multiple frames in the target sub-video are arranged in time sequence to form a to-be-processed sequence {t-n,..t-1, t, t+1,...t+m}. Then, the to-be-processed sequence is subjected to standardization processing, such as uniform size, graying and the like, so that the formats of all pre-processed images remain consistent. The selected key frame is used as a reference point, and the subsequent long short-term memory network learns the space-time pattern of the frames before and after the key frame.

[0068] Specifically, the image enhancement model comprises:

[0069] a convolutional neural network, configured to extract spatial features of the standardized sequence to obtain a feature sequence;

[0070] a long short-term memory network, configured to analyze the time sequence relationship between the key frame and the remaining frames in the feature sequence to obtain a time sequence analysis result;

[0071] a background frame determination module, configured to determine, as background frames, frames in the feature sequence that are of a preset continuous number and have a frame-to-frame pixel difference within a preset range;

[0072] an image denoising module, configured to perform denoising processing on the feature sequence according to the time sequence analysis result to obtain a denoised feature sequence, select a target frame from the continuous background frames every preset number of frames under the condition that the continuous background frames are subjected to displacement correction, determine a similar frame corresponding to the target frame, perform denoising processing on the target frame, and perform denoising processing on the similar frame corresponding to the target frame by using a denoising parameter of the target frame;

[0073] an image super-resolution module, configured to perform displacement correction on the denoised sequence feature according to the time sequence analysis result and a motion compensation algorithm to obtain an enhanced target sub-video, perform displacement correction on the target frame under the condition that the continuous background frames are subjected to displacement correction, and perform displacement correction on the similar frame corresponding to the target frame by using a displacement correction parameter of the target frame.

[0074] In practical applications, a convolutional neural network (CNN) is used to extract spatial features (such as edges and textures) of each frame of the target sub-video to obtain a feature sequence. Local feature processing can also be defined, such as extracting only the local picture of the personnel entering the monitoring video to improve the speed and accuracy of processing.

[0075] After obtaining the feature sequence, a long short-term memory network (LSTM) is used to process the time sequence features, avoiding the gradient disappearance problem of traditional RNN. The pre-processed feature sequence is input into the LSTM in chronological order, and each frame of feature is processed at each step. By adjusting the number of LSTM layers and hidden units, the inter-frame span that can be memorized (such as the dynamic information of the previous 10-30 frames) is controlled, and the feature values are used to determine the relevance of the current frame and the historical frame. For example, if an object moves to the right in 5 consecutive frames, it is inferred that the position of the object in the current frame should have a rightward displacement deviation from the previous frame.

[0076] The outdoor shooting monitoring video has random noise due to the change of sunlight angle over time and the wind blowing the camera, but the random noise between frames is correlated in continuous time. The change in time sequence obtained by comparing the pixel points of the key frame t with those of other frames can effectively remove the noise. For example, the noise and picture jitter offset in the t-1 frame are determined by comparing the pixel values of the t-1 frame and the t frame. For example, due to jitter, 20 rows of pixels are added to the t-1 frame (there are no 20 rows of pixels in the t frame). Therefore, the t-1 frame will be de-noised starting from the 21st row based on the characteristics of the key frame.

[0077] Since the monitoring video frames are continuously changing and strongly associated with each other, there is no frame jump like in movies or short videos. After identifying the key frame (i.e., the frame with clear and good brightness), the key frame can be used to compensate for the key motion object picture before and after it. For example, the motion character in the previous 3 frames is blurred, and the key frame (the 4th frame) can extract the features of the motion character to repair the blur in the previous 3 frames.

[0078] Since the monitoring video is shot in a certain area for a long time, the time proportion of static pictures is large, and the time proportion of dynamic pictures is small. The static pictures can be regarded as background frames to reduce the relevant calculations. Specifically, after extracting the background frames, a preset number of frames can be set to 3 frames, and a target frame is selected every 3 frames. The target frame is processed, and this processing is also applied to the similar frames corresponding to the target frame, such as the previous 3 frames. Dynamic pictures are processed, that is, frames other than background frames are processed once (such as character sharpening, noise removal, etc.), thereby improving the processing efficiency.

[0079] In practical applications, for the training of the image enhancement model, training samples are collected in advance, covering low-resolution images of various scenes and different degradation levels and their corresponding high-resolution reference images. By in-depth analysis of the degradation mechanism and causes of the degraded images, such as motion blur, noise interference, compression distortion, etc., a theoretical basis is provided for subsequent algorithm design. At the same time, the quality evaluation function and loss function of the reconstructed image, such as the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM), are analyzed and studied to measure the difference between the reconstructed image and the original high-definition image, guiding the parameter adjustment in the model training process. Until the image enhancement model converges, the trained image enhancement model is obtained. The trained image enhancement model is used to perform image enhancement processing on the target sub-video.

[0080] Step S3, receiving the enhanced target sub-video sent by each node, splicing all the enhanced target sub-videos to obtain an image enhanced video.

[0081] In practical applications, the enhanced target sub-videos sent by each node are spliced according to the time sequence. In the splicing process, a preset number of frames are expanded before and after the splicing point to obtain the expanded frame segment corresponding to the splicing point. For example, the first server processes frames 1-10, the second server processes frames 11-20, the last frame of the first server after processing is frame 10, and frame 11 processed by the second server. Frame 10 and frame 11 are taken as the splicing point. If the preset number of frames is 1, the expanded frame segment is frame 10 to frame 11. The mean value of frames 10 and 11 is adjusted to realize smooth processing. If the effect needs to be better, the preset number of frames can be increased, such as 2. The expanded frame segment is frame 9 to frame 12. The mean value of frames 9-12 is adjusted to realize smooth transition.

[0082] Specifically, after obtaining the image enhanced video, the following steps are included:

[0083] In the case where the frame rate of the image enhanced video is less than the preset frame rate, according to the time sequence analysis result and the number of dynamic pictures, a dynamic compensation algorithm is used to insert compensation frames to obtain a frame-inserted image enhanced video.

[0084] In practical applications, there are cases where the frame rate of the monitoring video picture is low, especially for old cameras, which have a frame rate of only 25 or 30 frames. According to the time sequence correlation between pictures obtained by LSTM analysis, compensation frames can be inserted in the enhanced picture according to the number of dynamic pictures with dynamic compensation effect, such as inserting three frames continuously for frames with large dynamic changes and compensating at multiple different positions, so as to increase the frame rate to 60 frames. On the basis of optimizing the picture, the video smoothness is improved.

[0085] The video image enhancement method provided by the application cuts the video according to pixel difference, concurrently transmits the target sub-video after cutting to the corresponding node, effectively reduces the dependence of the algorithm on hardware, and improves the processing speed of image enhancement. By taking pixel difference as the cutting basis, the key frame is selected, and then the remaining frames in the target sub-video are processed according to the key frame, and the response speed is fast.

[0086] As shown in Figure 2 The embodiment of the application further provides a video image enhancement system, which comprises:

[0087] A video cutting module 10 is configured to cut a video according to pixel difference to obtain a plurality of target sub-videos.

[0088] A video distribution module 20 is configured to distribute each target sub-video to a corresponding node, and the node performs image enhancement processing on the received target sub-video to obtain an enhanced target sub-video.

[0089] A receiving and splicing module 30 is configured to receive the enhanced target sub-video sent by each node, splice all the enhanced target sub-videos to obtain an image enhanced video.

[0090] The modules of the video image enhancement system can be realized by software, hardware or a combination thereof. The modules and units can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so that the processor can call and execute the operations of the modules.

[0091] As shown in Figure 3 The embodiment of the application discloses a computer device, which comprises a memory and a processor, and the memory stores a computer program.

[0092] The computer device can be a server, and its internal structure diagram can be as shown in Figure 3 The computer device comprises a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide calculation and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the video image enhancement method described in the above embodiments.

[0093] Those skilled in the art can understand that Figure 3The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0094] The embodiment of the present application also discloses a computer readable storage medium storing a computer program, wherein the computer program causes a computer to execute the video image enhancement method described in each of the above embodiments.

[0095] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, the processes of the above-mentioned embodiments can be included. Any reference to memory, storage, database or other medium used in each embodiment provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0096] The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not contradict, they should be considered as the scope of the present application.

Claims

1. A video image enhancement method, characterized in that, include: The video is segmented based on pixel differences to obtain multiple target sub-videos; Each of the target sub-videos is assigned to a corresponding node, and the node performs image enhancement processing on the received target sub-video to obtain an enhanced target sub-video. The system receives the enhanced target sub-videos sent by each of the nodes, and stitches all the enhanced target sub-videos together to obtain the image-enhanced video.

2. The video image enhancement method as described in claim 1, characterized in that, The process of segmenting the video based on pixel differences to obtain multiple target sub-videos includes: Determine the starting segmentation point of the video, and determine the pixel difference calculation point based on the starting segmentation point; Calculate the pixel difference between adjacent frames in the video starting from the pixel difference calculation point; Each pixel difference is compared with a preset threshold to obtain the correlation between adjacent frames corresponding to each pixel difference; The video is segmented based on the correlation between each adjacent frame to obtain multiple target sub-videos.

3. The video image enhancement method as described in claim 2, characterized in that, Determining the starting segmentation point of the video and determining the pixel difference calculation point based on the starting segmentation point includes: The average pixel difference within a preset number of frames is calculated starting from the first frame of the video; If the average pixel difference is greater than a threshold, the next frame after the candidate frame is taken as the starting frame. The candidate frame is the last frame within a preset number of frames. The step of calculating the average pixel difference within a preset number of frames starting from the starting frame of the video is executed. If not, the starting frame is taken as the starting segmentation point. The preset number of frames is added to the starting segmentation point to obtain the pixel difference calculation point.

4. The video image enhancement method as described in claim 2, characterized in that, The step of performing image enhancement processing on the received target sub-video to obtain the enhanced target sub-video includes: Determine the keyframes of the target sub-video; The frames of the target sub-video after the keyframe is determined are arranged in time sequence to obtain the sequence to be processed. The sequence to be processed is then standardized to obtain the standardized sequence. The standardized sequence is input into the trained image enhancement model to obtain video enhancement frames.

5. The video image enhancement method as described in claim 4, characterized in that, The image enhancement model includes: A convolutional neural network is used to extract spatial features from the standardized sequence to obtain a feature sequence. A long short-term memory network is used to analyze the temporal relationship between key frames and other frames in the feature sequence to obtain temporal analysis results; The background frame determination module is used to select a preset number of consecutive frames in the feature sequence, and the inter-frame pixel difference is within a preset range, as background frames. The image denoising module is used to denoise the feature sequence according to the time series analysis results to obtain a denoised feature sequence. Under the condition of displacement correction of continuous background frames, a target frame is selected from the continuous background frames every preset number of frames, the corresponding close frame is determined, the target frame is denoised, and the denoising parameters of the target frame are used to denoise the corresponding close frame. The image super-resolution module is used to perform displacement correction on the denoised sequence features based on the temporal analysis results and motion compensation algorithm to obtain an enhanced target sub-video. While performing displacement correction on continuous background frames, displacement correction is performed on the target frame, and the displacement correction parameters of the target frame are used to perform displacement correction on the adjacent frames corresponding to the target frame.

6. The video image enhancement method as described in claim 5, characterized in that, After obtaining the image-enhanced video, the process includes: If the frame rate of the image-enhanced video is less than the preset frame rate, a dynamic compensation algorithm is used to insert compensation frames based on the timing analysis results and the number of dynamic frames to obtain the image-enhanced video with the added frames.

7. The video image enhancement method as described in claim 1, characterized in that, The process of stitching together all the enhanced target sub-videos to obtain the image-enhanced video includes: All enhanced target sub-videos are stitched together according to the time sequence. During the stitching process, a preset number of frames are extended before and after the stitching point as the center to obtain the extended frame segment corresponding to the stitching point. The extended frame segments are smoothed to obtain an image-enhanced video.

8. A video image enhancement system, characterized in that, include: The video segmentation module is used to segment a video into multiple target sub-videos based on pixel differences; The video allocation module is used to allocate each of the target sub-videos to the corresponding nodes, and the nodes perform image enhancement processing on the received target sub-videos to obtain enhanced target sub-videos. The receiving and stitching module is used to receive the enhanced target sub-videos sent by each of the nodes, and stitch all the enhanced target sub-videos together to obtain the image-enhanced video.

9. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

Images