Video processing method and video processing system

Abstract

The invention provides a video processing method and a video processing system, and belongs to the technical field of video processing, the method comprises the following steps: in response to a received first parameter, executing conversion processing on a currently input first video to obtain a second video; the first parameter is used for indicating a conversion object subjected to conversion processing in the first video, and the conversion object at least comprises at least one of a frame rate, a code rate, a resolution ratio, a scanning mode and noise; and on the basis of a received second parameter, performing detail enhancement processing on a to-be-played video frame in a currently played third video, and playing the video frame after the detail enhancement processing, the third video comprising the first video and / or the second video, wherein the detail enhancement processing is used for enhancing textures in a plurality of texture images with different scales decomposed from the video frame to be played, and the second parameter is used for adjusting the enhancement degree of the texture images.

Description

Technical Field

[0001] The present disclosure relates to the technical field of video processing, and in particular to a video processing method and a video processing system. Background Art

[0002] With the development of display technology, large-size display screens are gradually being widely used in homes, shopping malls, stations and other occasions. For example, it is often necessary to publish video sources to multiple display devices for playback. However, the models and types of multiple display devices may be different. Some display devices can play videos with higher definition, such as 4K display screens that can play 4K videos. Some display devices are affected by hardware performance, and the clarity and resolution of the videos they play are not high. In related technologies, videos published from the publishing end to multiple display devices often cannot adapt to the display performance of each display device. Summary of the invention

[0003] Based on the content of the background technology, the present disclosure proposes a video processing method and a video processing system.

[0004] In a first aspect of the present disclosure, a video processing method is provided, the method comprising:

[0005] In response to the received first parameter, a conversion process is performed on the currently input first video to obtain a second video; the first parameter is used to indicate a conversion object in the first video to be converted by the conversion process, and the conversion object includes at least one of a frame rate, a bit rate, a resolution, a scanning mode and noise; and,

[0006] Based on the received second parameter, performing detail enhancement processing on a video frame to be played in a third video currently being played, and playing the video frame after the detail enhancement processing, wherein the third video includes the first video and / or the second video;

[0007] The detail enhancement process is used to enhance the texture in a plurality of texture images of different scales decomposed from the video frame to be played, and the second parameter is used to adjust the degree of enhancement of the texture image.

[0008] Exemplarily, the conversion object includes a scanning mode, the first video is an interlaced scanning mode, and the performing of the conversion process on the currently input first video in response to the received first parameter includes:

[0009] For an i-th video frame and an i+1-th video frame adjacent to each other in the first video, generating a missing video frame based on the i-th video frame and the i+1-th video frame, wherein the missing video frame represents an intermediate picture of the movement of the i-th video frame to the i+1-th video frame;

[0010] The i-th video frame or the i+1-th video frame is inserted into the missing video frame to convert the scanning mode of the first video into a line-by-line scanning mode.

[0011] Exemplarily, the conversion object includes a frame rate, and the performing conversion processing on the currently input first video in response to the received first parameter includes:

[0012] Obtaining a first frame rate of the first video and a second frame rate in the first parameter;

[0013] Based on the first frame rate and the second frame rate, determining a plurality of target video frames to be inserted in the first video, and a number of inserted frames corresponding to each of the target video frames;

[0014] For each of the target video frames, based on the target video frame and a subsequent video frame adjacent to the target video frame and having a subsequent playback sequence number, generate the number of predicted video frames as many as the number of interpolated frames;

[0015] Insert the number of predicted video frames into between the target video frame and the subsequent video frame to convert the frame rate of the first video.

[0016] Exemplarily, before generating the number of predicted video frames for each target video frame based on the target video frame and a subsequent video frame adjacent to the target video frame and having a subsequent playback sequence number, the method further includes:

[0017] Determine whether the first video is converted from an interlaced mode to a progressive scan mode before the current one;

[0018] If not, the scanning mode of the first video is converted into the progressive scanning mode, and after being converted into the progressive scanning mode, the frame rate of the converted first video is converted.

[0019] Exemplarily, the conversion object includes noise, and the performing conversion processing on the currently input first video in response to the received first parameter includes:

[0020] Extracting at least one video group from the first video, wherein the video group includes a plurality of adjacent original video frames;

[0021] Based on the multiple original video frames in the video group, establishing a position correspondence relationship between pixels belonging to the same content in different video frames;

[0022] Based on the position correspondence, screening out noise candidate regions from the original video frame;

[0023] Determine, based on the pixel values ​​of the pixels of each original video frame of the noise candidate region in the video group, an evaluation value corresponding to the noise candidate region, wherein the evaluation value is used to characterize the confidence that the noise candidate region is noise;

[0024] Based on the evaluation value, determining a target area from a plurality of noise candidate areas;

[0025] The target area in each video frame of the first video is subjected to denoising processing.

[0026] Exemplarily, the performing detail enhancement processing on the video frame to be played in the third video currently being played based on the received second parameter includes:

[0027] Decomposing the video frame to be played in the third video N times to obtain a structural image and N texture images; wherein, in the nth decomposition, the structural image obtained in the n-1th decomposition is decomposed, and the nth texture image is obtained according to the difference between the structural image obtained in the n-1th decomposition and the structural image obtained in the nth decomposition; wherein N is a positive integer greater than 1, and n is a positive integer greater than 0 and less than or equal to N;

[0028] Based on the second parameter and the structure image, respectively perform texture enhancement on the N texture images;

[0029] The N texture images after texture enhancement are fused with the structural image.

[0030] Exemplarily, when decomposing the video frame to be played in the third video each time, the process includes:

[0031] Get the decomposition operator corresponding to the current decomposition, as well as the sliding window and step size corresponding to the current decomposition;

[0032] Based on the decomposition operator, taking the diagonal of the sliding window as the center line, weighted averaging is performed according to the step size for each pixel in the corresponding first image located within the sliding window;

[0033] According to the result of the weighted average, obtaining the corresponding second image;

[0034] Based on the difference between the first image and the second image, obtaining a corresponding texture image;

[0035] The sizes of the sliding windows corresponding to the N decompositions are the same or not completely the same, and the decomposition operators corresponding to the N decompositions are the same or not completely the same.

[0036] Exemplarily, the decomposition operator includes a spatial domain weight and a range weight for performing the weighted average; wherein the spatial domain weight and the range weight are obtained by the following steps:

[0037] Generating different weight groups of the decomposition operator, each of the weight groups comprising a generated preset spatial domain weight and a preset range weight;

[0038] For each of the weight groups, based on the weight group, at least one original image is decomposed N times to obtain a decomposed image group corresponding to the weight group, wherein the decomposed image group includes a predicted structure image and N predicted texture images;

[0039] Based on the predicted structure image, texture enhancement is performed on the N predicted texture images respectively, and the enhanced multiple predicted texture images are fused with the predicted structure image to obtain a fused image corresponding to each weight group;

[0040] Acquire evaluation values ​​corresponding to a plurality of the fused images, wherein the evaluation values ​​are used to evaluate the picture quality of the fused images;

[0041] Based on each of the evaluation values, selecting a target weight group from the plurality of weight groups;

[0042] The preset spatial domain weight and the preset range weight in the target weight group are used as the spatial domain weight and the range weight in the decomposition operator respectively.

[0043] Exemplarily, the decomposition operators corresponding to the N decompositions are not exactly the same;

[0044] The weight group includes m sub-weight groups, each of which corresponds to a decomposition operator used for at least one decomposition in the N decompositions, and m is less than or equal to N.

[0045] Exemplarily, the decomposition operator includes a spatial domain weight and a range weight for performing the weighted average; wherein the spatial domain weight and the range weight are obtained by the following steps:

[0046] Acquire multiple sample image groups, wherein the sample image groups include original image samples and target image samples, wherein the target image samples are obtained by performing texture enhancement processing on the original image samples, wherein the processing method adopted by the texture enhancement processing includes: at least one of a global optimization method and a sub-window variance filter based on a local statistical edge model;

[0047] Inputting the original image samples in the plurality of sample image groups into a target neural network to obtain predicted image samples output by the target neural network, wherein a decomposition operator is preset in the target neural network;

[0048] According to the difference between the target image sample and the predicted image sample, the spatial domain weight and the range weight of the decomposition operator preset in the target neural network are updated to obtain the weighted average spatial domain weight and the range weight.

[0049] Exemplarily, the target neural network is used to decompose the original image sample, and to perform texture enhancement on the structure image and texture image obtained after the decomposition, and the detail enhancement processing is performed on the video frame to be played in the third video currently being played based on the received second parameter, including:

[0050] Inputting the second parameter and the video frame to be played into the trained target neural network;

[0051] Obtain a video frame output by the trained target neural network.

[0052] Exemplarily, the performing texture enhancement on each of the N texture images based on the second parameter and the structure image includes:

[0053] For the nth texture image, based on the second parameter, obtaining a first weight corresponding to the nth texture image and a second weight corresponding to the structure image; wherein the second parameter is used to indicate the size and / or category of the first weight;

[0054] fusing the nth texture image and the structural image according to the first weight and the second weight to obtain a mapping image corresponding to the nth texture image;

[0055] The N texture images after texture enhancement are fused with the structural image, including:

[0056] The N mapping images and the structural image are fused.

[0057] Exemplarily, the first weight is a fixed value, or the first weight is a nonlinearly transformed value.

[0058] A second aspect of the present disclosure provides a video processing system, including a server and a terminal, wherein the terminal is used to play a video sent by the server; wherein:

[0059] The server is configured to provide an interactive interface, and based on a first parameter input in the interactive interface, perform conversion processing on the first video to obtain a second video; wherein the first parameter is used to indicate a conversion object in the first video to be converted, and the conversion object includes at least one of a frame rate, a bit rate, a resolution, a scanning mode, and noise;

[0060] The terminal is configured with a video player, and is configured to perform detail enhancement processing on the video frame to be played of the third video based on the input second parameter, and send the enhanced video frame to the video player for playing;

[0061] The detail enhancement processing is used to enhance the texture in a plurality of texture images of different scales decomposed from the third video, and the second parameter is used to adjust the degree of enhancement of the texture image.

[0062] Exemplarily, the terminal further includes a graphics processor GPU, and the video player runs on the GPU;

[0063] The GPU is configured to perform detail enhancement processing on the video frames to be played in the video player based on the input second parameter, and send the enhanced video frames to the display screen of the terminal for playback.

[0064] By adopting the video processing method provided by this embodiment, the first video currently input can be converted in response to the received first parameter to obtain the second video; and based on the received second parameter, the video frame to be played in the third video currently played is enhanced in detail, and the video frame after the detail enhancement is played, the third video includes the first video and / or the second video; wherein the first parameter is used to indicate the conversion object to be converted in the first video, and the conversion object at least includes at least one of the frame rate, bit rate, resolution, scanning mode and noise; the detail enhancement is used to enhance the texture in the texture images of multiple different scales decomposed from the third video, and the second parameter is used to adjust the degree of enhancement of the texture image. Since the first video can be converted according to the first parameter, at least one of the frame rate, bit rate, resolution, scanning mode and noise of the first video can be converted into a type adapted to the display device, thereby adapting to different display devices; in this way, it is not necessary to prepare video sources of corresponding formats for different display devices separately. Secondly, the first video or the third video can also be enhanced in detail in response to the second parameter, thereby enhancing the texture details of the third video, which helps to improve the clarity of the video.

[0065] The above description is only an overview of the technical solution of the present disclosure. In order to more clearly understand the technical means of the present disclosure, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present disclosure more obvious and easy to understand, the specific implementation methods of the present disclosure are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or related technologies, the following is a brief introduction to the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the proportions in the drawings are only for illustration and do not represent the actual proportions.

[0067] Figure 1 The system architecture diagram of the video processing system is shown;

[0068] Figure 2 A schematic diagram of the software and hardware deployment of a video processing system is shown;

[0069] Figure 3 A schematic diagram of an interactive interface is shown;

[0070] Figure 4 A schematic diagram of an interactive interface on a terminal is shown;

[0071] Figure 5 A schematic diagram showing the deployment of a terminal for detail enhancement processing is shown;

[0072] Figure 6 A video processing method proposed in this embodiment is exemplarily described;

[0073] Figure 7 A schematic diagram of decomposing an original image is exemplarily shown;

[0074] Figure 8 A schematic diagram showing image alignment is shown;

[0075] Fig. 9 A schematic diagram of the detail enhancement process is shown;

[0076] Fig.10 Two schematic diagrams of operations when performing decomposition are shown;

[0077] Fig.11 Schematic diagrams of two nonlinear mapping methods are shown;

[0078] Fig.12 A schematic diagram showing the deployment of decomposition operators according to method 2 is shown. DETAILED DESCRIPTION

[0079] In order to make the above-mentioned purposes, features and advantages of the present disclosure more obvious and understandable, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0080] In the related art, with the continuous updating and iteration of display devices, some display devices have the ability to play high-definition videos, such as 4K videos. However, 4K video sources cannot keep up with the demand at present, and there is a problem of insufficient 4K sources, which cannot provide sufficient 4K ultra-high-definition video services for users with high-definition display devices. In addition, there are many types of display devices. For example, the commonly used display devices are mobile phones, tablets, computers, etc. The hardware performance of mobile phones and computers is higher than that of tablets. Therefore, the picture quality of the same video played on a mobile phone and on a tablet may be different; in this way, when the same video is played on display devices with different hardware performance, there is also a problem of poor balance of picture quality of videos played on different display devices.

[0081] In view of this, an embodiment of the present disclosure proposes a video processing method, which can perform conversion processing on a currently input first video in response to a received first parameter to obtain a second video; and based on the received second parameter, perform detail enhancement processing on the video frames to be played in the currently played third video, and play the video frames after the detail enhancement processing, wherein the third video includes the first video and / or the second video; wherein the first parameter is used to indicate the conversion object to be converted in the first video, and the conversion object includes at least one of the frame rate, bit rate, resolution, scanning mode and noise; the detail enhancement processing is used to enhance the texture in multiple texture images of different scales decomposed from the third video, and the second parameter is used to adjust the degree of enhancement of the texture image.

[0082] Among them, the first parameter can be input by the user, so that according to the first parameter, the object to be converted in the first video can be converted, so that at least one of the frame rate, bit rate, resolution, scanning mode and noise of the first video can be converted into a type that is compatible with the display device, so that the same video source can be played on different types of display devices according to user needs, and comprehensive video services can be provided to users according to the type of display device and user needs. For example, the first video that is not of 4K type can be converted into a 4K video, so as to provide 4K video high-definition services for display devices capable of playing 4K videos, so that 4K videos can be supplied with existing video sources, and are no longer restricted by insufficient supply of 4K sources.

[0083] Among them, the details of the first video and / or the second video can also be enhanced in response to the second parameter, thereby enhancing the texture details of the third video, which helps to improve the clarity of the video. This can help to perform image processing on the texture in the video screen according to user needs during video playback, thereby improving the picture quality from the image processing dimension.

[0084] The video processing method and the video processing system of this embodiment are exemplarily described below with reference to the accompanying drawings.

[0085] Reference Figure 1 and Figure 2 As shown, in one embodiment, a video processing system is provided. Figure 1 The system architecture diagram of the video processing system is shown. Figure 2 The schematic diagram of the hardware and software deployment of the video processing system is shown in FIG. Figure 1 and Figure 2 As shown, the video processing system in this embodiment may include a server and a terminal, and the terminal is used to play the video sent by the server; wherein, it may include multiple terminals, and the multiple terminals can be connected to the server.

[0086] The server is configured to provide an interactive interface, and based on a first parameter input in the interactive interface, perform conversion processing on the first video to obtain a second video; wherein the first parameter is used to indicate a conversion object to be converted in the first video, and the conversion object includes at least one of a frame rate, a bit rate, a resolution, a scanning mode, and noise;

[0087] The terminal is configured with a video player, and is configured to perform detail enhancement processing on the video frame to be played of the third video based on the input second parameter, and send the enhanced video frame to the video player for playback;

[0088] The detail enhancement processing is used to enhance the texture in a plurality of texture images of different scales decomposed from the third video, and the second parameter is used to adjust the degree of enhancement of the texture image.

[0089] like Figure 1 and Figure 2 As shown, the video processing method of this embodiment can be applied to scenes such as video playback in public areas, radio and television playback, and conference video playback. For example, it can be applied to scenes such as radio and television, film and television, and cultural and museum.

[0090] like Figure 1 As shown, the types of the various terminals may be the same or different, and the terminals have display screens, and the sizes of the display screens may be different. For example, the various terminals may include 4K display screens, tablets, mobile phones, large display screens, integrated display machines, televisions, conference tablets, commercial display advertising machines, spliced ​​screens, etc.

[0091] like Figure 1 and Figure 2 As shown, the server can be a server or located in the cloud, such as a public cloud, a private cloud, a hybrid cloud, etc.; or, the server can be Figure 2 The micro workstation and video encoding processing all-in-one machine shown in the figure.

[0092] The server can provide an interactive interface that supports human-computer interaction. Figure 3 As shown, Figure 3 A schematic diagram of an interactive interface is shown. The interactive interface may be a web interface of a video processing system. A plurality of selection controls may be set on the interactive interface. Different selection spaces may correspond to different conversion objects, such as Figure 3 As shown, multiple selection controls may include: frame doubling, deinterlacing, super-resolution, video noise reduction and other controls. When the user selects these controls, the first parameter will be automatically generated. In this way, the first parameter can be generated according to the conversion object corresponding to the selected selection control. For example, if the user selects "frame doubling, deinterlacing", the first parameter may include "frame doubling / deinterlacing", indicating that the first video needs to be deinterlaced and the frame rate of the first video needs to be converted.

[0093] It should be noted that if the user only selects "double frame" but does not select "deinterlacing", the server can automatically select "deinterlacing" because double frame must be performed under deinterlacing. Therefore, the server automatically generates the first parameter of "double frame / deinterlacing".

[0094] In this embodiment, the server determines which objects of the first video need to be converted based on the first parameter, thereby converting at least one conversion object of the first video in turn, and the conversion object may include at least one of frame rate, bit rate, resolution, scanning mode and noise.

[0095] In this embodiment, after the server completes the conversion of the first video, the second video obtained after the conversion can be sent to the terminal, and the terminal can play the second video sent. Of course, in some embodiments, if the first video does not need to be converted, the server can also directly send the first video to the terminal for playback, so that the video played on the terminal can be the first video or the second video.

[0096] Among them, the terminal can perform detail enhancement processing on the video picture of the video being played, so as to improve the picture clarity of the video being played as much as possible based on the existing hardware configuration of the terminal through data processing means. The video played on the terminal can be called a third video, and the third video can be the first video or the second video. Therefore, the object of the detail enhancement processing can be the first video or the second video.

[0097] The terminal may also provide an interactive interface, which may be customized to the video player on the terminal, see Figure 4 As shown, Figure 4 A schematic diagram of the interactive interface on the terminal is shown, Figure 4 As shown, an interactive interface can be provided in the playback interface of the video player, such as the interactive interface can be configured as an interactive window, the first control and the second control can be displayed in the interactive interface, the first control can be used to prompt the user whether to turn on detail enhancement processing, and the second control can be used to prompt the user to input a second parameter (i.e., the intensity of detail enhancement processing), such as providing three intensity levels of "high, medium, and low". In practice, the corresponding second parameter can be generated according to the user's operation on the first control in the interactive interface, such as touch operation, and the operation on the second control, such as input operation.

[0098] By adopting the video processing system of this embodiment, the server can convert the first video to be played in advance so that the converted video can adapt to the hardware performance of the terminal, so that the smoothness, resolution, clarity, etc. of the video played on the terminal are more adapted to the terminal, thereby improving the quality of video playback; on the terminal, it is also possible to respond to the second parameter and perform detail enhancement processing on the video being played, which helps to perform image processing on the texture in the video picture according to user needs during the video playback process, thereby improving the picture quality from the image processing dimension.

[0099] In a video processing system provided in an embodiment, the terminal is also configured with a graphics processor GPU. Figure 5 As shown, Figure 5 FIG. 4 shows a schematic diagram of the deployment of the terminal for detail enhancement processing, such as Figure 5 As shown, the terminal also includes a graphics processor GPU, and the video player runs on the GPU;

[0100] The GPU is configured to perform detail enhancement processing on the video frames to be played in the video player based on the input second parameter, and send the enhanced video frames to the display screen of the terminal for playback.

[0101] In the video processing system of this embodiment, the terminal is equipped with a GPU, and the GPU can be used to perform detail enhancement processing on the video frames to be played of the third video, and send the enhanced video frames to the video player for playback; in this way, the third video can be processed using the existing hardware of the terminal itself, so that the terminal uses the GPU of the SoC to process the video without increasing the hardware cost across platforms. Therefore, it can run on various terminals and PC GPU architectures, and the terminal in this embodiment supports Android, iOS, Windows, Linux, Mac and other systems.

[0102] The video player can be any existing type of player, such as Figure 5 As shown, the video player can be a MediaPlayer player, and the MediaPlayer player can run on a GPU, so that the GPU can extract the video frames to be played from the cache area of ​​the video player, and then perform detail enhancement processing on the video frames. The video frames after detail enhancement processing replace the video frames to be played in the cache area, and enable the player to play the video frames after detail enhancement processing on the display screen.

[0103] Below, combined with the attached Figure 6 A video processing method proposed in this embodiment is exemplarily described as follows: Figure 6 As shown, the video processing method may include the following steps:

[0104] Step S101: in response to the received first parameter, performing conversion processing on the currently input first video to obtain a second video;

[0105] The first parameter is used to indicate a conversion object to be converted in the first video, and the conversion object includes at least one of a frame rate, a bit rate, a resolution, a scanning mode and noise.

[0106] Step S102: Based on the received second parameter, detail enhancement processing is performed on the video frames to be played in the third video currently being played, and the video frames after the detail enhancement processing are played, where the third video includes the first video and / or the second video;

[0107] The detail enhancement processing is used to enhance the texture in a plurality of texture images of different scales decomposed from the third video, and the second parameter is used to adjust the degree of enhancement of the texture image.

[0108] In this embodiment, the first video can be uploaded to the server by the user. For example, the user can provide the storage address of the first video in the interactive interface, and the server can download or retrieve the first video from the storage address. For another example, the user can import the first video to the server.

[0109] Among them, the first parameter can be input by the user on the interactive interface provided by the server, such as generated after selecting multiple selection controls, which represents the conversion object that the user wants to convert for the first video. The first parameter can include the name of the conversion object, such as frame rate, bit rate, resolution, scanning mode, or include the conversion operation to be performed on the conversion object. For example, for the corresponding frame rate, its first parameter can be "double frame", indicating that the frame rate of the first video needs to be increased during the conversion process; for example, for the resolution, its first parameter can be "super-resolution", indicating that the resolution of the first video needs to be increased during the conversion process.

[0110] Among them, the frame rate refers to the number of video frames played in a unit of time, such as 1 second. The bit rate refers to the size of the data encoded by the encoder per second. The resolution refers to the number of pixels per inch in the image, which is related to the clarity of the image. The scanning methods include progressive scanning and interlaced scanning. The progressive scanning method refers to the method of scanning the image continuously line by line by an electron beam; the interlaced scanning method is completed by two-field scanning. The first field (odd field) only scans the odd lines, and the second field (even field) only scans the even lines. Interlaced scanning technology plays a great role in the case of insufficient transmission signal bandwidth.

[0111] Among them, the first parameter may include not only the conversion object, but also the parameters of the converted second video such as frame rate, bit rate, resolution, scanning mode, etc., that is, the first parameter may indicate not only the conversion object to be converted, but also the specifications of the converted second video. In this way, when converting the first video, the first video may be converted into a second video that meets the video specifications indicated by the first parameter, that is, the second video meets the video specifications indicated by the first parameter, such as the resolution, frame rate, bit rate, scanning mode, etc. indicated by the first parameter.

[0112] When converting the frame rate of the first video, the scanning mode of the first video needs to be converted to a progressive scanning mode. The conversion process can be called de-interlacing, and de-interlacing can be performed in an inter-field de-interlacing mode, a single-field de-interlacing mode, or the like.

[0113] In this embodiment, after obtaining the second video, the second video can be sent to the terminal, or the second video can be saved on the server. The first video can be played on the terminal, and the second video can also be played on the server. The video being played on the terminal is called the third video.

[0114] In this embodiment, when playing the third video, the video frame of the third video can be image processed to improve the playback quality of the third video at the pixel granularity. Specifically, as described above, the terminal can receive the second parameter input by the user, and perform detail enhancement processing on the video frame to be played in the third video according to the second parameter, and the detail enhancement processing is mainly used to enhance the texture in the video frame to be played, such as enhancing the contrast of the texture, sharpening the texture, and appropriately thickening the texture.

[0115] The process of detail addition processing may be: first decomposing the video frame to be played into a structural image and a texture image, then enhancing the texture image, and combining the structural image to form a new video frame. When the texture image is enhanced, the degree of enhancement of the texture image may be adjusted according to the second parameter, for example, the contrast and sharpness of the texture image may be adjusted.

[0116] In this embodiment, the texture image mainly focuses on the texture performance of the local area in the image, describing the grayscale or color change law between pixels, and the structural image mainly focuses on the structural features in the image, such as edge features, corner features and contour features. Among them, the structural image and multiple texture images can be decomposed from the video frame to be played. The specific decomposition process can be: the video frame to be played (hereinafter referred to as the original image) is first decomposed to decompose the structural image, and the structural image is subtracted from the original image to obtain the texture image; then, the structural image is decomposed again (second decomposition) to obtain the second structural image, and then the second structural image is subtracted from the structural image obtained by the first decomposition to obtain the second decomposed texture image, and so on, multiple texture images and multiple structural images can be obtained.

[0117] Among them, since multiple texture images are obtained, and each texture image is obtained by the difference between the structural images obtained during two adjacent decompositions, the multiple texture images can be understood as texture images of different scales, and the different scales can be reflected in the grayscale, color change and receptive field of the decomposed texture image.

[0118] For example, refer to Figure 7 As shown, Figure 7 The schematic diagram of decomposing the original image is exemplified. It can be seen that a structural image and multiple texture images are obtained. Different texture images have different details. Generally, as the decomposition progresses, the detail expression of the texture image is enhanced.

[0119] In this embodiment, the first video can be converted into a second video that meets the user's needs based on the first parameter. In this way, the second video can adapt to the hardware performance of the terminal used by the user, thereby achieving the display performance of the terminal and the adaptability of the played video. For example, if the terminal is a 4K display screen, the first video that is not in 4K format can be converted into a second video in 4K format, thereby achieving high-definition playback. On the one hand, it solves the problem of insufficient high-definition 4K video sources, and on the other hand, it can improve the adaptability between the played video and the display performance of the display device, thereby providing users with diverse video services.

[0120] In this embodiment, according to the second parameter, detail enhancement processing can be performed on the video frame to be played in the third video currently being played. Since detail enhancement processing is image processing performed on the picture in the video frame, the video frame can have a clearer texture by processing the pixel points in the video frame, thereby improving the picture quality of the video being played from the perspective of image processing, and can improve the playback quality of the video being played as much as possible when the display performance of the terminal is insufficient.

[0121] In some embodiments, the conversion of the scanning mode can convert the interlaced mode of the first video into a progressive scanning mode, and this process can be called deinterlacing. The deinterlacing method includes inter-field deinterlacing. Inter-field deinterlacing is a method of combining consecutive fields into one field. In simple terms, it is to merge the pixel rows of the upper field (odd field) and the lower field (even field) to obtain a complete frame. However, when the picture moves, such as the movement of an object between the upper field and the lower field, the picture cannot be completely matched, and a picture effect similar to "ghosting" or "jaggies" will appear.

[0122] Therefore, in this embodiment, a motion compensation de-interlacing method is adopted. This method can perform dynamic prediction based on two adjacent fields, predict the movement mode of the moving object between the two fields, generate the missing field information, and then add the information of the upper or lower field to the field information supplemented by the motion estimation to generate a complete frame of the picture, thereby improving "ghosting" or "jaggies".

[0123] Therefore, when the first video is in interlaced scanning mode and the conversion object includes scanning mode, the first video needs to be deinterlaced. In specific implementation, a missing video frame can be generated based on the i-th video frame and the i+1-th video frame for the adjacent i-th video frame and the i+1-th video frame in the first video; and the i-th video frame or the i+1-th video frame is inserted into the missing video frame to convert the scanning mode of the first video into a progressive scanning mode; wherein the missing video frame represents the intermediate picture of the movement of the i-th video frame to the i+1-th video frame.

[0124] In this embodiment, the i-th video frame and the i+1-th video frame can be respectively referred to as the upper field and the lower field in the first video, that is, the two adjacent fields described above. In practice, an optical flow algorithm can be used to predict the motion information of an object in a video frame between the i-th video frame and the i+1-th video frame, thereby obtaining the missing motion information between the i-th video frame and the i+1-th video frame, and a missing video frame can be generated based on the missing motion information.

[0125] In this way, based on the changes in pixels in the i-th video frame and the i+1-th video frame in the time domain and the correlation between the i-th video frame and the i+1-th video frame, the motion information of the object between the i-th video frame and the i+1-th video frame can be calculated by finding the correspondence between the i-th video frame and the i+1-th video frame.

[0126] In this embodiment, the generated missing video frame can be used to make up for the missing information of the upper field or the lower field. Depending on whether the i-th video frame or the i+1-th video frame is to be inserted, the missing video frame can be the upper field information or the lower field information. If the i-th video frame (upper field) is to be inserted, the missing video frame can be the lower field information, so that the missing video frame can be used as the even-numbered lines in a complete frame, and the i-th video frame can be the odd-numbered lines in a complete frame. If the i+1-th video frame (lower field) is to be inserted, the missing video frame can be the upper field information, so that the missing video frame can be used as the odd-numbered lines in a complete frame, and the i-th video frame can be the even-numbered lines in a complete frame.

[0127] Exemplarily, for the i-th video frame, the i+1-th video frame, and the i+2-th video frame, missing video frame 1 is generated between the i-th video frame and the i+1-th video frame, and missing video frame 1 and the i-th video frame are combined to form a complete video frame. For missing video frame 2 is generated between the i+1-th video frame and the i+2-th video frame, missing video frame 2 and the i+1-th video frame are combined to form a complete video frame, and so on. Each video frame in the first video is deinterlaced.

[0128] Thus, in every two i-th video frames and i+1-th video frames, a missing video frame is generated based on the correlation between them, and the i-th video frame or the i+1-th video frame is inserted into the missing video frame to obtain a complete frame of video, so that each video frame can be deinterlaced through the missing video frame, thereby performing deinterlacing with dynamic compensation between the upper and lower fields. And because the generated missing video frame performs motion estimation based on the correlation between the upper and lower fields, the motion gap between the missing video frame and the inserted i-th video frame or the i+1-th video frame is reduced, thereby improving the "ghosting" or "jaggies" phenomenon.

[0129] In some embodiments, the frame rate of the first video may be changed to change the smoothness of playing the first video. For example, the frame rate of the first video may be increased to improve the smoothness of playing the first video; or, for another example, the frame rate of the first video may be decreased to reduce the smoothness of playing the first video.

[0130] In this embodiment, taking increasing the frame rate of the first video as an example, the process of performing conversion processing on the currently input first video may include the following steps:

[0131] Step S201: Obtain a first frame rate of a first video and a second frame rate in a first parameter;

[0132] Step S202: determining a plurality of target video frames to be inserted in the first video and the number of inserted frames corresponding to each target video frame based on the first frame rate and the second frame rate;

[0133] Step S203: for each target video frame, based on the target video frame and a subsequent video frame adjacent to the target video frame and having a subsequent playback sequence number, generate an insertion frame number of predicted video frames;

[0134] Step S204: inserting the predicted video frames having the number of inserted frames between the target video frame and the next video frame to convert the frame rate of the first video.

[0135] Among them, the second frame rate can be specified by the user. For example, the frame rate increase multiples can be provided in the interactive interface, such as one time, two times, three times, four times, etc. According to the user's choice of the increase multiple, the second frame rate in the first parameter is generated. If the user chooses to increase it by one time, the second frame rate can be twice the first frame rate.

[0136] Wherein, the second frame rate is greater than the first frame rate, and the multiple target video frames to be interpolated and the number of interpolated frames corresponding to each target video frame can be determined according to the ratio between the second frame rate and the first frame rate. Wherein, the multiple target video frames can be each video frame in the first video, and the number of interpolated frames can be greater than or equal to 1. Wherein, the multiple target video frames can be some video frames in the first video frame, and the number of interpolated frames can be greater than or equal to 1. For example, the 2nth video frame in the first video is used as the target video frame, and n can be 1, 2, 3, etc., then the even-numbered frames in the first video are used as the target video frames.

[0137] Among them, the number of interpolated frames corresponding to each target video frame can be determined by the first frame rate, the second frame rate and the number of target video frames. For example, the target video frame is each video frame in the first video, the first frame rate is 25fps, and the second frame rate is 50fps, then the number of interpolated frames corresponding to each target video frame is 1; for another example, the target video frame is an even-numbered video frame in the first video, the first frame rate is 25fps, and the second frame rate is 50fps, then the number of interpolated frames corresponding to each target video frame is 2.

[0138] In this example, the video frame inserted after each target video frame can be predicted based on the motion relationship between the target video frame and video frames adjacent to the target video frame in the first video. The predicted video frame can be called a predicted video frame.

[0139] In a specific implementation, an optical flow algorithm can be used to calculate the motion relationship between the target video frame and a subsequent video frame that is adjacent to the target video frame and has a subsequent playback sequence number, thereby generating an inserted number of predicted video frames. The predicted video frames can be used to supplement the missing motion information of the moving object between the target video frame and the subsequent video frame, so that the target video frame, the subsequent video frame and the predicted video frame inserted between the target video frame and the subsequent video frame form a continuous motion picture, which can improve the smoothness between pictures and the smoothness of picture changes.

[0140] When the number of predicted video frames inserted after the target video frame is greater than 1, such as when two predicted videos are to be inserted, in one example, predicted video frame 1 can be generated based on the target video frame and the next video frame, and then predicted video frame 2 can be generated based on the correlation between predicted video frame 1 and the target video frame, and predicted video frame 3 can be generated based on the correlation between predicted video frame 1 and the next video frame, and so on. Multiple predicted video frames can be generated based on the generated predicted video frames and combined with the next video frame after the target video frame. The number of generated predicted video frames is greater than the number of interpolated frames, such as the difference between the number of interpolated frames and the number of interpolated frames can be 1, that is, one more predicted video frame is generated.

[0141] For example, assuming that the number of inserted frames is 2, three predicted video frames are generated, as described above, namely predicted video frame 1, predicted video frame 2, and predicted video frame 3, and then predicted video frame 2 and predicted video frame 3 can be inserted after the target video frame. Alternatively, the picture similarity between predicted video frame 1, predicted video frame 2, and predicted video frame 3 can be compared, and then the two predicted video frames with higher similarity can be inserted after the target video frame.

[0142] Of course, in some other examples, interpolation can also be performed by directly copying video frames. For example, for the i-th frame video, the i-th frame video can be copied and the copied i-th frame video can be inserted between the i-th frame video and the i+1-th frame video.

[0143] Accordingly, as described above, before converting the frame rate of the first video, the scanning mode of the first video needs to be converted into a line-by-line scanning mode, so that each frame of the first video can have complete frame information, thereby predicting the motion information between adjacent frames and generating a predicted video frame.

[0144] Specifically, when the first parameter includes the frame rate, it can be determined whether the first video has been converted from the interlaced scanning mode to the progressive scanning mode before the current one; if so, the frame rate conversion is performed according to the above steps S201-S204; if not, the scanning mode of the first video is converted to the progressive scanning mode, and after the conversion to the progressive scanning mode, the frame rate of the converted first video is converted.

[0145] The process of converting to the progressive scanning mode may refer to the process of the de-interlacing mode described above, and will not be described in detail here.

[0146] In some embodiments, the first parameter may also indicate processing of the noise in the first video, such as performing noise reduction processing, so that the first video is converted from a high noise ratio to a low noise ratio. Specifically, during the noise reduction processing, in order to improve the noise reduction effect, it may compare multiple adjacent video frames in the first video, determine the noise location, and then perform noise reduction on the location where the noise is located. The specific process can refer to the following process:

[0147] Step S301: extracting at least one video group from a first video, wherein the video group includes a plurality of adjacent original video frames;

[0148] Step S302: establishing a position correspondence relationship between pixels belonging to the same content in different video frames based on multiple original video frames in the video group;

[0149] Step S303: based on the position correspondence, filter out the noise candidate area from the original video frame;

[0150] Step S304: determining an evaluation value corresponding to the noise candidate region based on the pixel values ​​of the pixels of each original video frame in the video group, where the evaluation value is used to characterize the confidence that the noise candidate region is noise;

[0151] Step S305: determining a target area from a plurality of noise candidate areas based on the evaluation value;

[0152] Step S306: performing denoising processing on the area where the target area is located in each video frame of the first video.

[0153] In this embodiment, in order to increase the accuracy of noise location positioning, multiple groups of adjacent video frames in the first video can be compared, and the noise position can be determined based on the multiple groups of comparison results. Specifically, at least one video group can be extracted from the first video, and each video group includes multiple adjacent original video frames. Among them, one video group can be extracted, or multiple video groups can be extracted, and the number of original video frames included in each video frame can be 2, 3, 4, etc. The more the number of original video frames included, the more accurate the positioning can be, but it also brings about an increase in computational complexity. Therefore, multiple video groups can be selected, and each video group includes 2 or 3 original video frames for comparison.

[0154] For each video group, multiple original video frames in the video group may be aligned. The alignment refers to aligning pixels belonging to the same content in multiple original video frames, that is, establishing a position correspondence relationship between pixels of the same content in different original videos. Figure 8 As shown, a schematic diagram of image alignment is shown, such as Figure 8 As shown, the same content ( Figure 8 The target point A in the original video frame is at position (x1, y1) in the original video frame 1, and the target point B is at position (x2, y2) in the original video frame 2. The position correspondence is the correspondence between the position (x1, y1) and the position (x2, y2) of the target point A.

[0155] Thus, the position correspondence reflects the position change of the same content in different original video frames, so that the same content can be located from different original video frames. In this way, the region that may be a noise point can be determined from the original video frame through the position correspondence, and this region is called a noise point candidate region. For example, the pixel values ​​of the positions of the same content in different original video frames can be determined based on the position correspondence, and the region where the noise point appears can be determined based on the difference between the pixel values. For example, Figure 8 As shown in the figure, assuming that the pixel values ​​of target point A in original video frame 1 and original video frame 2 have a small difference, it can be determined that there is no noise at the position where target point A is located in original video frame 1 and original video frame 2; if the difference is large, it can be determined that there is noise at the position where target point A is located in original video frame 1 and original video frame 2. Then, by comparing the pixel value at the corresponding position (x1, y1) in original video frame 2 and the pixel value at the position (x2, y2) in original video frame 1, the noise occurrence area can be determined.

[0156] It should be noted that when multiple video groups are included, each video group obtains a set of noise candidate areas based on the alignment results (position correspondence). In this case, the multiple sets of noise candidate areas obtained from the multiple video groups can be deduplicated, so that the screened noise candidate areas include all the noise in the first video as much as possible.

[0157] In this way, the area where the noise occurs in the first video can be determined, and then the area where the noise actually occurs can be determined from the noise candidate area, avoiding the problem of misjudgment of noise due to light intensity difference, occlusion difference, etc. In this embodiment, the evaluation value corresponding to the noise candidate area can be determined based on the pixel value of the pixel point of each original video frame in the video group of the noise candidate area. The evaluation value is used to characterize the confidence that the noise candidate area is noise. Specifically, the pixel value corresponding to the noise candidate area can reflect the brightness of the area, that is, the noise intensity. The greater the noise intensity, the higher the probability that it is noise.

[0158] In practice, a pixel value threshold may be preset, the pixel value corresponding to the noise candidate area may be compared with the pixel value threshold, and the evaluation value may be determined according to the comparison result.

[0159] Among them, each noise point candidate region can be sorted according to the size of the evaluation value, and according to the sorting result, the target region is screened out, and the target region is the region where the noise point is located. For example, the larger the evaluation value, the higher the probability of the representation being a noise point, and the evaluation values ​​can be sorted from large to small, and then the top m noise point candidate regions can be used as the target region.

[0160] The target area in each video frame in the first video may be subjected to denoising, and the denoising is used to eliminate the noise in the target area, and the denoising may be performed by using a mean filter, an adaptive Wiener filter or a median filter.

[0161] By aligning the multiple video frames proposed in this embodiment to locate the positions of the noise points, the accuracy of noise point positioning can be improved, thereby improving the noise reduction effect.

[0162] In some embodiments, when performing detail enhancement processing on the third video, the video frame to be played in the third video may be first decomposed into a structural image and a texture image, and then the texture image may be enhanced in detail. Fig. 9 As shown, a schematic diagram of the detail enhancement process is shown. Fig. 9 As shown, the specific process can refer to the following steps:

[0163] Step S401: Decomposing the video frame to be played in the third video N times to obtain a structural image and N texture images; wherein, in the nth decomposition, the structural image obtained in the n-1th decomposition is decomposed, and the nth texture image is obtained according to the difference between the structural image obtained in the n-1th decomposition and the structural image obtained in the nth decomposition; wherein N is a positive integer greater than 1, and n is a positive integer greater than 0 and less than or equal to N;

[0164] Step S402: Based on the second parameter and the structure image, texture enhancement is performed on each of the N texture images;

[0165] Step S403: Fusing the N texture images and the structural image after texture enhancement.

[0166] In this embodiment, the video frame to be played may be decomposed N times, and the decomposition process has been described above and will not be repeated here. The structural image may be the structural image obtained by the first decomposition.

[0167] Among them, Fig. 9 As shown, at the nth decomposition, the structural image obtained by the n-1th decomposition can be decomposed according to the following formula (1):

[0168] U i =f i (U i-1 ) formula (1);

[0169] In formula (1), U i represents the nth structural image, U i-1 represents the n-1th structural image, f i Represents a decomposition operator.

[0170] Among them, Fig. 9 As shown, at the nth decomposition, the nth texture image can be obtained according to the following formula (2):

[0171] D i =U i -U i-1 Formula (2);

[0172] In formula (2), U i represents the nth structural image, U i-1 represents the n-1th structural image, D i Represents the nth texture image.

[0173] In this embodiment, Fig. 9As shown, when N texture images are obtained, since different texture images are obtained from different decompositions of the structural image, the N texture images may have different texture representations, which may reflect the texture details in the video frame from different scales.

[0174] Among them, Fig. 9 As shown, the texture image can be subjected to detail operation according to the second parameter, and the detail operation can refer to: based on the second parameter and the structure image, each of the N texture images is subjected to texture enhancement, and the texture enhancement can adjust the contrast, sharpness, etc. of the texture image according to the second parameter. Different texture images can correspond to different enhancement degrees, such as corresponding to different contrast and sharpness adjustments.

[0175] Among them, when the texture image is texture enhanced, it can be combined with the structural image. For example, when the texture image is adjusted according to the second parameter, the information of the structural image can be integrated to ensure that the overall structure (contour) of the texture image remains unchanged when the texture image is texture enhanced.

[0176] In some embodiments, each texture image may correspond to a texture-enhanced image, and N texture-enhanced texture images may be obtained. The N texture-enhanced texture images and the structural image may be fused to complete detail enhancement processing of the video frame to be played.

[0177] In an exemplary embodiment of this embodiment, when the decomposition operator is used to decompose the video frame to be played, the decomposition operator can use a diagonal weighted average method when performing the decomposition. Specifically, when the video frame to be played in the third video is decomposed each time, the following steps can be followed:

[0178] Step S501: Obtain the decomposition operator corresponding to the current decomposition, as well as the sliding window and step size corresponding to the current decomposition;

[0179] Step S502: Based on the decomposition operator, taking the diagonal line of the sliding window as the center line, weighted averaging is performed on each pixel in the first image corresponding to the current time and located within the sliding window according to the step length;

[0180] Step S503: acquiring the corresponding second image according to the weighted average result;

[0181] Step S504: obtaining a corresponding texture image based on the difference between the first image and the second image;

[0182] The sizes of the sliding windows corresponding to the N decompositions are the same or not completely the same, and the decomposition operators corresponding to the N decompositions are the same or not completely the same.

[0183] In this embodiment, the decomposition operator may be the decomposition operator f in the above formula (1): i， The sliding window refers to a window used for weighted averaging of the image to be decomposed (structural image or video frame to be played), and the step size may refer to the step size of weighted averaging within the sliding window, which may be 1 or 2.

[0184] If it is the first decomposition, the currently corresponding first image may be a video frame to be played; if it is not the first decomposition, the currently corresponding first image may be a structural image obtained by the last decomposition.

[0185] Please refer to Fig.10 As shown, Fig.10 Two schematic diagrams of operations for decomposition are shown, such as Fig.10 As shown, for the current corresponding first image, based on the decomposition operator, the diagonal line in the sliding window is used as the center line, and the weighted average of each sub-pixel in the sliding window is performed in sequence according to the step size. Fig.10 As shown in (a) and (b), the pixels on the diagonal of the sliding window in the first image need to be weighted averaged, where the size of the sliding window in (a) is smaller than that in (b), so that (a) has a smaller receptive field. The step size of the sliding window in (a) is 1, indicating that the weighted average is performed with each pixel on the diagonal as the center, and the step size of the sliding window in (b) is 2, indicating that the weighted average needs to be performed with every other pixel on the diagonal as the center.

[0186] In this embodiment, the decomposition operator includes weight values ​​required for weighted averaging, such as spatial domain weights and range weights.

[0187] In this embodiment, the first image can be subjected to a weighted average of diagonal lines according to a sliding window, so that texture information can be effectively extracted. The result of the weighted average can be regarded as a decomposition of the first image to obtain a corresponding second image. If the first image is the original video frame, the second image is a structural image. If the first image is a structural image, the second image is also a structural image.

[0188] The first image and the second image can be subtracted according to the above formula (2) to obtain the corresponding texture image.

[0189] Each decomposition requires a decomposition operator. In practice, the decomposition operators used in each decomposition may be consistent, thereby ensuring the consistency of N decompositions. Of course, in some embodiments, when the decomposition operators corresponding to each decomposition are the same, the weight values ​​included in the decomposition operators may not be completely the same, such as including different range weights and spatial weights, so that decompositions of different strengths can be performed.

[0190] In some examples, the sizes of sliding windows corresponding to the N decompositions may be the same, so that the sizes of the obtained texture images can be consistent, which is convenient for later fusion. In some other examples, the sizes of sliding windows corresponding to the N decompositions may not be exactly the same, so that texture images under different receptive fields can be obtained. In this way, when enhancing each texture image, enhancements of different scales can be performed, and when fusing each texture image, texture features under different receptive fields can be fused, which helps to improve the detail expression of the video frame.

[0191] In some examples of this embodiment, the decomposition operator includes a spatial domain weight and a range weight for weighted averaging, wherein the spatial domain weight measures the clustering between two pixel points, and the farther the distance, the lower the weight, and the range weight measures the similarity of the pixel values ​​between the two pixel points, and the more similar, the greater the weight. Among them, one of the two pixel points can be a pixel point on the diagonal line, and the other can be a pixel point other than the pixel point on the diagonal line.

[0192] In this embodiment, the decomposition operator is related to the decomposition scale of the video frame to be played, and its multiple texture images are related to the subsequent texture enhancement effect. Therefore, in order to achieve a better texture enhancement effect, it is necessary to construct a suitable decomposition operator, for example, it is necessary to obtain the optimal spatial domain weight and range weight in the decomposition operator. In this embodiment, two methods of obtaining a better decomposition operator are proposed, namely, method 1 and method 2.

[0193] Method 1, multiple weight groups can be generated in advance, for example, multiple decomposition operators can be generated in advance, and different decomposition operators contain different spatial domain weights and range weights. In this way, the same image can be decomposed by multiple decomposition operators respectively, and finally an image with detail enhancement is obtained. By comparing the image quality of the image with detail enhancement, the optimal decomposition operator can be screened, that is, the optimal spatial domain weight and range weight can be obtained.

[0194] Please refer to the following steps for details:

[0195] Step S601: generating different weight groups of decomposition operators, each weight group including generated preset spatial domain weights and preset range weights;

[0196] Step S602: for each weight group, based on the weight group, at least one original image is decomposed N times to obtain a decomposed image group corresponding to the weight group, wherein the decomposed image group includes a predicted structure image and N predicted texture images;

[0197] Step S603: Based on the predicted structure image, texture enhancement is performed on each of the N predicted texture images, and the enhanced multiple predicted texture images and the predicted structure image are fused to obtain a fused image corresponding to each weight group;

[0198] Step S603: obtaining evaluation values ​​corresponding to a plurality of fused images, where the evaluation values ​​are used to evaluate the picture quality of the fused images;

[0199] Step S604: based on each evaluation value, select a target weight group from multiple weight groups;

[0200] Step S605: using the preset spatial domain weight and the preset range weight in the target weight group as the spatial domain weight and the range weight in the decomposition operator respectively.

[0201] In this embodiment, one weight group may correspond to one decomposition operator, which can be regarded as generating multiple decomposition operators, each weight group including a preset spatial domain weight and a preset value domain weight. When generating multiple weight groups, they can be generated randomly or according to a certain weight spacing.

[0202] In one example, when generating multiple weight groups, a set of functions can be selected as spatial domain and range weighting functions, such as Gaussian, Huber, Laplace, Cauchy, Welsch, Charbonnier, etc. Then, the parameters of these functions are generated in batches, and the parameter settings are set according to a certain interval, thereby obtaining multiple weight groups.

[0203] The decomposition operator corresponding to each weight group may be used to decompose the same original image N times, or multiple original images may be decomposed N times. For example, the decomposition operator containing the same weight group may decompose multiple original images N times respectively.

[0204] Among them, after decomposition, each original image and each weight group corresponds to a decomposed image group, such as original image 1 and weight group 1 correspond to decomposed image group 1, and decomposed image group 1 includes a predicted structure image and N predicted texture images obtained by decomposing the original image 1 N times using weight group 1.

[0205] The predicted texture images in each image decomposition group can be texture enhanced as described above. It should be noted that the predicted texture images in different image decomposition groups correspond to the same texture enhancement. Then, the enhanced multiple predicted texture images and predicted structure images can be fused to obtain a fused image with enhanced details. In this way, each image decomposition group corresponds to a fused image, that is, each original image and each weight group corresponds to a fused image.

[0206] Among them, the picture quality corresponding to the multiple fused images can be evaluated manually or by machine to obtain the evaluation values ​​corresponding to each of the multiple fused images, and the evaluation values ​​are used to evaluate the picture quality of the fused image. In one example, a reference-free image quality evaluation index based on machine learning / deep learning can be calculated for the fused image. The index is divided into two categories: perceptual image quality distortion evaluation and natural scene image aesthetic quality evaluation. Taking the aesthetic quality of natural scene images as an evaluation index as an example, it can simulate the evaluation of enhanced quality by three types of evaluators, including ordinary consumers, image and video enthusiasts, practitioners and experts.

[0207] Among them, based on each evaluation value, the image quality of each fused image can be sorted from high to low, and the fused image with the highest evaluation value can be screened out. The weight group corresponding to the fused image with the highest evaluation value can be used as the spatial domain weight and range weight in the decomposition operator.

[0208] In an example of this embodiment, combined with the above, different decompositions can correspond to decomposition operators containing different weight values. Therefore, when generating weight groups, m sub-weight groups can be generated in batches each time, and the m sub-weight groups are combined into one weight group to correspond to the N decomposition operators (i.e., N weight values) used in the N decompositions of an image.

[0209] Accordingly, each sub-weight group corresponds to a decomposition operator used in at least one of the N decompositions, where m is less than or equal to N.

[0210] In this example, among the m sub-weight groups in the same weight group, at least one of the range weight and the spatial weight in different sub-weight groups is different. m can be less than N, in which case, at least two decompositions in the N decompositions use the same sub-weight group. Among them, m can be equal to N, in which case, each decomposition in the N decompositions uses a different sub-weight group.

[0211] Method 2: Use machine learning to train the optimal decomposition operator. The specific process is as follows:

[0212] Step S701: obtaining a plurality of sample image groups, the sample image groups comprising original image samples and target image samples, the target image samples being obtained by texture enhancement processing of the original image samples, the processing method adopted for texture enhancement processing comprising: at least one of a global optimization method and a sub-window variance filter based on a local statistical edge model;

[0213] Step S702: inputting original image samples in the plurality of sample image groups into a target neural network to obtain predicted image samples output by the target neural network, wherein a decomposition operator is preset in the target neural network;

[0214] Step S703: According to the difference between the target image sample and the predicted image sample, the spatial domain weight and the range weight of the decomposition operator preset in the target neural network are updated to obtain the weighted average spatial domain weight and the range weight.

[0215] In this embodiment, each sample image group may include original image samples and target image samples. Different sample image groups include original image samples of different contents. The target image samples are obtained by performing texture enhancement processing on the original image samples. The processing methods adopted may include global optimization methods, sub-window variance filtering based on local statistical edge models, etc., wherein multiple sample image groups may correspond to multiple texture enhancement processing methods, for example, including sample image groups obtained by processing using global optimization methods, and also including sample image groups obtained by sub-window variance filtering based on local statistical edge models.

[0216] It should be noted that, in addition to the above two methods, a non-real-time detail enhancement method can be used, which has better effects, can be enhanced at a large rate, is not prone to artifacts (such as gradient flipping, ringing, halo) or is not suitable for GPU parallelization.

[0217] In this embodiment, the sample image group can be used as a training sample to train the target neural network, and the target neural network can be used to decompose the input image, and after performing texture enhancement on each texture image obtained by the decomposition, the structural image and the enhanced texture image are merged to output a predicted image sample. Specifically, the original image sample can be input into the target neural network, and the target neural network can perform detail enhancement processing on the original image sample according to the process of steps S401-S402 above to output a predicted image sample.

[0218] Next, a loss function can be constructed based on the predicted image samples and the target image samples, and the parameters in the target neural network can be updated based on the value of the loss function, specifically, including updating the weight values ​​(spatial weights and range weights) of the decomposition operators in the target neural network. In one example, the loss function can be constructed based on the similarity between the predicted image samples and the target image samples.

[0219] The decomposition operator to be updated in the target neural network is called a preset decomposition operator. After multiple updates, the spatial domain weight and the range weight in the preset decomposition operator are updated, and the decomposition operator at the end of training is finally extracted as the decomposition operator in step S401. In one example, the end of training can be conditioned on the similarity between the predicted image sample and the target image sample being less than a preset value.

[0220] By adopting this example, the decomposition operator can be directly obtained by machine learning, thereby improving the adaptability of the decomposition operator to various images.

[0221] In an example of the present embodiment, since the target neural network already has the ability to perform detail enhancement processing on images, the trained target neural network can be deployed on the terminal, and the video frame to be played and the second parameter can be directly input into the trained target neural network; then, the video frame output by the trained target neural network can be obtained.

[0222] Among them, the second parameter can be used to replace the parameters for performing texture enhancement on N texture images in the target neural network.

[0223] Thus, this method requires the terminal to have the hardware conditions for deploying a neural network. In one example, the decomposition operator obtained by the above method 1 and method 2 can be directly written into the GPU in the terminal, so that the GPU can decompose and enhance the texture of the incoming video frame according to the decomposition operator. Because the above steps S401-S403 are adopted, the computational complexity is the lowest, the hardware resources consumed are the least, and the hardware performance requirements are also the lowest, so that it can realize the simultaneous playback and processing of images on low-performance terminals.

[0224] In some embodiments, in step S402, it is necessary to perform texture enhancement on N texture images based on the second parameter and the structure image. This process may be called image mapping. In one example, mapping may be implemented according to the following process:

[0225] First, for the nth texture image, based on the second parameter, a first weight corresponding to the nth texture image and a second weight corresponding to the structure image are obtained; wherein the second parameter is used to indicate the size and / or category of the first weight;

[0226] Next, the nth texture image and the structural image are fused according to the first weight and the second weight to obtain a mapping image corresponding to the nth texture image;

[0227] After that, the N texture images and the structural images after texture enhancement are fused, including:

[0228] Next, the N mapping images and the structural image are fused.

[0229] In this embodiment, the mapping may be performed according to the following formula (3):

[0230] F1= U i + D i * k Formula (3)

[0231] F1 is the mapping image, k is the first weight, and the second weight is 1.

[0232] The method of formula (3) is a linear mapping method, that is, the texture image can be texture enhanced by using a linear mapping method.

[0233] Then, the N mapping images and the structural image can be fused to obtain a video frame with enhanced details. In one example, in addition to linear mapping, a nonlinear mapping method can also be used, so that noise and compression artifacts can be suppressed and details can be improved. Fig.11 As shown, two nonlinear mapping modes are shown, in which the first weight and the second weight vary nonlinearly. For example, the first weight is a nonlinearly varying value, and in one example, the second weight may also be a nonlinearly varying value.

[0234] Please refer to Fig.12 As shown, Fig.12 A schematic diagram of the deployment of decomposition operators according to method 2 is shown, as Fig.12 As shown, firstly, a board performance evaluation is performed, where the board refers to a board on a device used to run the method in the above-mentioned mode 2 to obtain a decomposition operator.

[0235] Next, for method 2, the network structure of the target neural network can be designed, and each initial weight in the network structure can be generated, including the initial weight of the decomposition operator. After that, the constructed target neural network is exported and converted into a fragment shader; after that, the target neural network is deployed in the webGL video player, and then the target neural network is tested in the Android system (Android) for speed testing. If the running speed meets the requirements, the subsequent model training is entered. After the model training is completed, the trained target neural network and the weights in the target neural network are exported and converted into a fragment source shader, and then deployed to the video player on the terminal. The speed and effect verification are performed on the terminal side (such as a personal computer or mobile phone). The effect verification can be used to judge the image quality of the fused image output by the target neural network. If the effect verification is not ideal, it returns to the model training again, and goes through the training and deployment process again until the effect meets expectations.

[0236] After verification is in accordance with expectations, the decomposition operator can be extracted and written into the GPU of the terminal, and an SDK interface that can call the decomposition operator can be set between the video player and the GPU, thereby implementing the above-mentioned detail enhancement processing steps.

[0237] In some embodiments, the processing of the first video is exemplarily described as follows:

[0238] S1. Video import:

[0239] 1) Video file input

[0240] 2) Support single file import and selected folder import;

[0241] 3) Support input of files containing Chinese characters and punctuation supported by the system;

[0242] S2. Select the save location:

[0243] 1) Click to select the final save location;

[0244] 2) The save location cannot contain a file with the same name as an existing file.

[0245] S3, video processing process:

[0246] 1) Packaging format: mp4 and mov.

[0247] 2) Video encoding: mp4 supports H.265 and H.264 encoding;

[0248] mov supports ProRes 422 and ProRes 422HQ;

[0249] 3) Encoding bit rate: The bit rate can be set in mp4 packaging format;

[0250] There is no restriction for mov (the encoding format determines the bit rate).

[0251] 4) Algorithm functions: De-interlacing, two-fold interpolation, two-fold super-resolution, and video noise reduction algorithms can be selected and combined. Among them, two-fold super-resolution is a method of changing the picture from a low-resolution image to a high-resolution image, supporting up-conversion of 1080P images to 4K resolution images. Super-resolution is more about filling and completing pixel content, and information filling is achieved through traditional algorithms such as interpolation algorithms and degradation models. The interpolation algorithm predicts and fills information through surrounding pixel information. The degradation model assumes that the low-resolution image is obtained by noise compression of the high-resolution image. By continuously denoising and retaining the effective low-resolution picture information, high-resolution information can be obtained.

[0252] Among them, during the processing, the file being processed can be displayed, and there is a button to terminate the processing;

[0253] During the processing, the processing speed and remaining time can be displayed; there is feedback when the processing is completed.

[0254] Next, you can export the video:

[0255] 1) Export video files;

[0256] 2) Support the export of files containing Chinese characters and system-supported punctuation;

[0257] 3) After the video is generated, the status is displayed on the interface;

[0258] During the processing, the interactive interface has the following functions:

[0259] 1) The window can be full screen, and the interface font will be enlarged synchronously after full screen

[0260] 2) After the software is closed, the background processing will be terminated.

[0261] The main process is:

[0262] ①After the user opens the system, click Add Video to import the original video to be processed; the user can select a single video file or select a folder. The user will batch process the selected single video file or video files in the folder.

[0263] ② The user needs to select the packaging format and encoding format required for the final exported video. The packaging formats include MP4 and MOV. The choice of packaging format also determines the choice of encoding format to a certain extent; MP4 encoding format is mainly H.264 / H.265 encoding format; MOV. format uses Apple's QuickTime technology, so it also supports ProRes422 and ProRes 422HQ;

[0264] ③Users can customize the video bit rate, ranging from 0 to 120M; if no setting operation is performed, the default video bit rate is 50M

[0265] ④ Processing function selection: De-interlacing. That is, convert the original interlaced video to progressive video, and select whether to give top field priority (odd lines) or bottom field priority (even lines).

[0266] ⑤ Double interpolation can only be used when video deinterlacing is turned on. This function can fill the imported video with 2 times the frame rate. At present, most high-definition resolution videos are 25fps; exported videos can reach 50fps.

[0267] ⑥ "Double Super Resolution" is an independent option and has no relationship with other options. If "Double Super Resolution" is required, turn the switch to the ON state. If "Double Super Resolution" is not required, turn the switch to the OFF state.

[0268] ⑦"HDR" is an independent option and has no relationship with other options. If you need to convert SDR video to HDR, turn the switch to ON. The final output video will be HDR video. At present, HDR video can only be viewed on monitors that support HDR display, otherwise there will be color and brightness deviations.

[0269] ⑧If all AI functions are turned off and the input video is output, the exported video will only be re-encoded and decoded according to the selected export video packaging format and encoding format.

[0270] 1.1 Table 1 shows the input image requirements

[0271] Table 1.1

[0272]

[0273] 1.2 Table 1 shows the input packaging and encoding format requirements

[0274] Table 1.2

[0275] Packaging format Image Format Encoding format OP-Atom MXF HD50I AVC High Profile MOV HD50I ProRes 422, ProRes 422HQ

[0276] 1.3 Table 1 shows the output image requirements

[0277] Table 1.3

[0278]

[0279] 1.4 Table 1 shows the requirements for output packaging and encoding formats.

[0280] Table 1.3

[0281] Packaging format Image Format Encoding format Chroma sampling format MP4 HD 50P 4K UHD 50P H.264 (AVC), H.265 (HEVC) 4:2:0 MOV HD 50P 4K UHD 50P ProRes 422, ProRes 422HQ 4:2:2

[0282] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0283] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0284] The above is a detailed introduction to a video processing method and system provided by the present disclosure. Specific examples are used in this article to illustrate the principles and implementation methods of the present disclosure. The description of the above embodiments is only used to help understand the method of the present disclosure and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present disclosure, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present disclosure.

[0285] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0286] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

[0287] References herein to "one embodiment," "embodiment," or "one or more embodiments" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present disclosure. In addition, please note that examples of the term "in one embodiment" do not necessarily all refer to the same embodiment.

[0288] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present disclosure can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0289] In the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present disclosure may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names.

[0290] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A video processing method, characterized in that: The method comprises: In response to the received first parameter, a conversion process is performed on the currently input first video to obtain a second video; the first parameter is used to indicate a conversion object in the first video to be converted by the conversion process, and the conversion object includes at least one of a frame rate, a bit rate, a resolution, a scanning mode and noise; and, Based on the received second parameter, performing detail enhancement processing on a video frame to be played in a third video currently being played, and playing the video frame after the detail enhancement processing, wherein the third video includes the first video and / or the second video; The detail enhancement process is used to enhance the texture in a plurality of texture images of different scales decomposed from the video frame to be played, and the second parameter is used to adjust the degree of enhancement of the texture image.

2. The video processing method according to claim 1, characterized in that: The conversion object includes a scanning mode, the first video is an interlaced scanning mode, and the conversion process is performed on the currently input first video in response to the received first parameter, including: For an i-th video frame and an i+1-th video frame adjacent to each other in the first video, generating a missing video frame based on the i-th video frame and the i+1-th video frame, wherein the missing video frame represents an intermediate picture of the movement of the i-th video frame to the i+1-th video frame; The missing video frame is inserted into the i-th video frame or the i+1-th video frame to convert the scanning mode of the first video into a line-by-line scanning mode.

3. The video processing method according to claim 1, characterized in that: The conversion object includes a frame rate, and the step of performing conversion processing on the currently input first video in response to the received first parameter includes: Obtaining a first frame rate of the first video and a second frame rate in the first parameter; Based on the first frame rate and the second frame rate, determining a plurality of target video frames to be inserted in the first video, and a number of inserted frames corresponding to each of the target video frames; For each of the target video frames, based on the target video frame and a subsequent video frame adjacent to the target video frame and having a subsequent playback sequence number, generate the number of predicted video frames as many as the number of interpolated frames; Insert the number of predicted video frames into between the target video frame and the subsequent video frame to convert the frame rate of the first video.

4. The video processing method according to claim 3, characterized in that: Before generating the number of predicted video frames for each target video frame based on the target video frame and a subsequent video frame adjacent to the target video frame and having a subsequent playback sequence number, the method further includes: Determine whether the first video is converted from an interlaced mode to a progressive scan mode before the current one; If not, the scanning mode of the first video is converted into the progressive scanning mode, and after being converted into the progressive scanning mode, the frame rate of the converted first video is converted.

5. The video processing method according to claim 1, characterized in that: The conversion object includes noise, and the step of performing conversion processing on the currently input first video in response to the received first parameter includes: Extracting at least one video group from the first video, wherein the video group includes a plurality of adjacent original video frames; Based on the multiple original video frames in the video group, establishing a position correspondence relationship between pixels belonging to the same content in different video frames; Based on the position correspondence, screening out noise candidate regions from the original video frame; Determine, based on the pixel values ​​of the pixels of each original video frame of the noise candidate region in the video group, an evaluation value corresponding to the noise candidate region, wherein the evaluation value is used to characterize the confidence that the noise candidate region is noise; Based on the evaluation value, determining a target area from a plurality of noise candidate areas; The target area in each video frame of the first video is subjected to denoising processing.

6. The video processing method according to claim 1, characterized in that: The performing detail enhancement processing on the video frame to be played in the currently played third video based on the received second parameter includes: Decomposing the video frame to be played in the third video N times to obtain a structural image and N texture images; wherein, in the nth decomposition, the structural image obtained in the n-1th decomposition is decomposed, and the nth texture image is obtained according to the difference between the structural image obtained in the n-1th decomposition and the structural image obtained in the nth decomposition; wherein N is a positive integer greater than 1, and n is a positive integer greater than 0 and less than or equal to N; Based on the second parameter and the structure image, respectively perform texture enhancement on the N texture images; The N texture images after texture enhancement are fused with the structural image.

7. The video processing method according to claim 6, characterized in that: Each time the video frame to be played in the third video is decomposed, the method includes: Get the decomposition operator corresponding to the current decomposition, as well as the sliding window and step size corresponding to the current decomposition; Based on the decomposition operator, taking the diagonal of the sliding window as the center line, weighted averaging is performed according to the step size for each pixel in the corresponding first image located within the sliding window; According to the result of the weighted average, obtaining the corresponding second image; Based on the difference between the first image and the second image, obtaining a corresponding texture image; The sizes of the sliding windows corresponding to the N decompositions are the same or not completely the same, and the decomposition operators corresponding to the N decompositions are the same or not completely the same.

8. The video processing method according to claim 7, characterized in that: The decomposition operator includes a spatial domain weight and a range weight for performing the weighted average; wherein the spatial domain weight and the range weight are obtained by the following steps: Generating different weight groups of the decomposition operator, each of the weight groups comprising a generated preset spatial domain weight and a preset range weight; For each of the weight groups, based on the weight group, at least one original image is decomposed N times to obtain a decomposed image group corresponding to the weight group, wherein the decomposed image group includes a predicted structure image and N predicted texture images; Based on the predicted structure image, texture enhancement is performed on the N predicted texture images respectively, and the enhanced multiple predicted texture images are fused with the predicted structure image to obtain a fused image corresponding to each weight group; Acquire evaluation values ​​corresponding to a plurality of the fused images, wherein the evaluation values ​​are used to evaluate the picture quality of the fused images; Based on each of the evaluation values, selecting a target weight group from the plurality of weight groups; The preset spatial domain weight and the preset range weight in the target weight group are used as the spatial domain weight and the range weight in the decomposition operator respectively.

9. The video processing method according to claim 8, characterized in that: The decomposition operators corresponding to the N decompositions are not exactly the same; The weight group includes m sub-weight groups, each of which corresponds to a decomposition operator used for at least one decomposition in the N decompositions, and m is less than or equal to N.

10. The video processing method according to claim 7, characterized in that: The decomposition operator includes a spatial domain weight and a range weight for performing the weighted average; wherein the spatial domain weight and the range weight are obtained by the following steps: Acquire multiple sample image groups, wherein the sample image groups include original image samples and target image samples, wherein the target image samples are obtained by performing texture enhancement processing on the original image samples, wherein the processing method adopted by the texture enhancement processing includes: at least one of a global optimization method and a sub-window variance filter based on a local statistical edge model; Inputting the original image samples in the plurality of sample image groups into a target neural network to obtain predicted image samples output by the target neural network, wherein a decomposition operator is preset in the target neural network; According to the difference between the target image sample and the predicted image sample, the spatial domain weight and the range weight of the decomposition operator preset in the target neural network are updated to obtain the weighted average spatial domain weight and the range weight.

11. The video processing method according to claim 10, characterized in that: The target neural network is used to decompose the original image sample and to perform texture enhancement on the structure image and texture image obtained after the decomposition, and the detail enhancement processing is performed on the video frame to be played in the third video currently being played based on the received second parameter, including: Inputting the second parameter and the video frame to be played into the trained target neural network; Obtain a video frame output by the trained target neural network.

12. The video processing method according to claim 6, characterized in that: The performing texture enhancement on the N texture images respectively based on the second parameter and the structure image comprises: For the nth texture image, based on the second parameter, obtaining a first weight corresponding to the nth texture image and a second weight corresponding to the structure image; wherein the second parameter is used to indicate the size and / or category of the first weight; fusing the nth texture image and the structural image according to the first weight and the second weight to obtain a mapping image corresponding to the nth texture image; The N texture images after texture enhancement are fused with the structural image, including: The N mapping images and the structural image are fused.

13. The video processing method according to claim 12, characterized in that: The first weight is a fixed value, or the first weight is a value of a nonlinear transformation.

14. A video processing system, characterized in that: It includes a server and a terminal, wherein the terminal is used to play the video sent by the server; wherein, The server is configured to provide an interactive interface, and based on a first parameter input in the interactive interface, perform conversion processing on the first video to obtain a second video; wherein the first parameter is used to indicate a conversion object in the first video to be converted, and the conversion object includes at least one of a frame rate, a bit rate, a resolution, a scanning mode, and noise; The terminal is configured with a video player, and is configured to perform detail enhancement processing on the video frame to be played of the third video based on the input second parameter, and send the enhanced video frame to the video player for playing; The detail enhancement processing is used to enhance the texture in a plurality of texture images of different scales decomposed from the third video, and the second parameter is used to adjust the degree of enhancement of the texture image.

15. The video processing system according to claim 14, characterized in that: The terminal further includes a graphics processor GPU, and the video player runs on the GPU; The GPU is configured to perform detail enhancement processing on the video frames to be played in the video player based on the input second parameter, and send the enhanced video frames to the display screen of the terminal for playback.

Images