VVC fast CU division method and storage medium based on space-time characteristics

By combining the time domain and airspace correlation of video frames, using SAD and texture differences to determine the direction of CU division, the problem of high VVC encoding complexity is solved, and fast CU division is realized, suitable for efficient video encoding and real-time communication scenarios.

CN115278260BActive Publication Date: 2024-11-19GUANGZHOU XINLING TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210836143.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-15
Publication Date
2024-11-19
Estimated Expiration
2042-07-15

AI Technical Summary

Technical Problem

The computational complexity of the VVC encoding standard is too high, resulting in a significant increase in encoding time. In particular, the CU division method of quad-tree nested multi-type trees is one of the main reasons, which affects its promotion and application.

Method used

A VVC fast CU division method based on space-time domain characteristics is proposed. By calculating the absolute error sum (SAD) of the current CU and the best matching CU, whether to continue division, and using texture differences to judge the direction of multi-forktree division, skip not The necessary division process combines the time domain and airspace correlation of video frames.

Benefits of technology

It significantly reduces the encoding complexity and time of VVC, improves encoding efficiency and stability, and is suitable for application scenarios with high real-time requirements such as video conferencing and video surveillance, and hardly reduces encoding quality and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115278260B_ABST
    Figure CN115278260B_ABST
Patent Text Reader

Abstract

The present invention claims a VVC fast CU division method and storage medium based on space-time domain characteristics, which belongs to the field of video coding and includes the following steps: using the video time domain correlation, calculating the absolute error sum (SAD) between the current CU and its best matching CU obtained after motion estimation, determining whether the current CU continues to be divided to the next depth by the size of the SAD value, and terminating all division processes of some CUs in advance; then, for the CU that needs to be divided to the next depth, using the spatial domain characteristics of the video frame, analyzing the texture difference of the sub-blocks in the horizontal and vertical directions of the current CU, judging its multi-tree division direction, and skipping the multi-tree division process in the horizontal or vertical direction in advance. The present invention can be used for coding under the VVC inter-frame coding configuration, significantly reducing the coding time of VVC without almost reducing the coding efficiency and coding quality, and can be applied to application scenarios such as video conferencing and video surveillance that have high requirements for real-time coding.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of video coding, and more specifically, provides a CU fast partitioning method suitable for universal video coding VVC (Versatile Video Coding), which can be applied to video coding scenarios with high real-time requirements such as video conferencing and surveillance video. Background Art

[0002] With the development of communication technology and video technology, people's demand for higher resolution videos is growing. Although high efficiency video coding (HEVC) has excellent coding performance, its coding efficiency is also stretched when faced with the huge amount of data brought by high-quality videos such as 4K and 8K. In order to solve this problem, in 2015, the Joint Video Exploration Team (JVET) began to formulate the next generation of video coding standards. In July 2020, the first version of the VVC standard was officially released.

[0003] The technical means of VVC are basically similar to those of previous generations of video coding standards. It is still a block-based hybrid coding framework, but almost every link of the coding has been improved. The new technologies added by VVC include: quadtree nested multi-type tree partitioning, 67 intra-frame prediction modes, cross-component linear model prediction (CCLM), position dependent intra prediction combination (PDPC), multiple reference line intra prediction (MRL), matrix weighted intra prediction (MIP) and intra sub-partitions (ISP). These new technologies enable VVC to continue to squeeze the information redundancy that HEVC has not yet eliminated, and achieve the requirement of doubling the overall coding efficiency. However, the addition of these new technologies also makes the computational complexity of VVC increase sharply. The most intuitive manifestation is that the coding time of VVC has increased significantly. The coding time of VVC has increased by more than 10 times compared with HEVC. Such a high coding complexity brings great difficulties to the promotion and practical application of the VVC standard. Therefore, how to significantly reduce the coding complexity of VVC while ensuring coding efficiency has become the focus of current research. The CU partitioning method of quadtree (QT) nested MTT (multi-type tree) adopted by VVC is one of the main reasons for the increase in coding complexity. Therefore, optimizing the CU partitioning structure and realizing fast CU partitioning are important research directions for VVC coding acceleration.

[0004] During the encoding process of VVC, each frame of the video is divided into several 128×128 coding tree units (CTUs), and then the CTUs are divided into CUs through QT and MTT. The CUs are further divided until the length or width of the block is equal to 4 or the maximum division depth is reached. MTT includes horizontal division and vertical division, and each direction includes binary tree (BT) division and ternary tree (TT) division. Therefore, a CU has at most five division modes: QT, horizontal binary tree (HBT), horizontal ternary tree (HTT), vertical binary tree (VBT) and vertical ternary tree (VTT). All possible division modes are placed in the candidate list. For each CU, the encoder needs to traverse all the division modes in the candidate list, calculate its rate-distortion cost, and then select the division mode with the smallest rate-distortion cost as the optimal division mode for the current CU. In the CU division process, if the division mode can be reduced, the encoding complexity can be significantly reduced and the video encoding speed can be accelerated. Therefore, under the premise of ensuring that the encoding quality and encoding efficiency are slightly reduced, how to skip unnecessary CU division modes and determine the best CU division mode in advance to significantly reduce the encoding complexity is a hot topic in the current VVC fast algorithm research.

[0005] CN112702599A, a VVC intra-frame fast coding method based on deep learning, comprising: obtaining a brightness coding block and a quantization parameter during the VVC coding process; inputting the brightness coding block and the quantization parameter into a deep learning model that considers asymmetric convolution and attention perception to obtain a prediction vector, using the prediction probability of each division mode as the optimal fast division mode as an element in the prediction vector; determining the reserved division mode based on the prediction vector; and performing fast division within the VVC frame based on the reserved division mode. Compared with the prior art, the present invention greatly reduces the coding complexity, improves the stability of the algorithm, and ensures the accuracy of the division result during the VVC intra-frame fast coding process. However, the fast algorithm proposed in the patent inputs the current brightness coding block and the quantization parameter into the deep learning model, which only utilizes the spatial domain correlation of the video frame, and does not consider the characteristic that the temporal domain correlation of adjacent coding frames is large. Therefore, the patent is only suitable for the intra-frame coding process of the video, and its application is limited.

[0006] The algorithm proposed in the present invention first uses the temporal correlation of the video to calculate the sum of absolute errors (SAD) between the current CU and its best matching CU obtained after motion estimation, and determines whether the current CU continues to be divided to the next depth by the size of the SAD value, and terminates all division processes of some CUs in advance; then, for the CU that needs to be divided to the next depth, the spatial domain characteristics of the video frame are used to analyze the texture differences of the sub-blocks in the horizontal and vertical directions of the current CU, determine its multi-tree division direction, and skip the multi-tree division process in the horizontal or vertical direction in advance. Therefore, the present invention makes full use of the temporal and spatial domain correlation of the video frame, effectively reduces the complexity of VVC coding, and can be widely used in video efficient compression and real-time communication scenarios, such as video conferencing, video monitoring and other applications with high real-time requirements. Summary of the invention

[0007] The present invention aims to solve the above problems of the prior art. A VVC fast CU division method based on space-time characteristics is proposed. The technical solution of the present invention is as follows:

[0008] A VVC fast CU division method based on space-time domain characteristics comprises the following steps:

[0009] S1. Obtain the coding information of the current coding unit CU to be divided in the coding frame, and determine whether the current CU can continue to be divided. If the current CU division depth is less than the set maximum division depth, continue to divide, add all feasible division modes of the current CU to the division list, and enter step S2; otherwise, enter step S6;

[0010] S2, determine the size and type of the current CU, if the current CU is a 128×128 or 64×64 brightness CU, go to step S3, otherwise, go to step S5;

[0011] S3, read the pixel value of the best matching CU obtained by motion estimation of the current CU in the reference frame, calculate the pixel difference value between the current CU and the best matching CU, if the pixel difference value is less than the threshold, go to step S6, otherwise go to step S4;

[0012] S4, respectively calculating the texture difference of the sub-block in the horizontal direction and the texture difference of the sub-block in the vertical direction of the current CU, and judging whether to skip the multi-tree partitioning process in a certain direction by comparing the sizes of the texture differences in the two directions, and then deleting the multi-tree partitioning mode in the partitioning list;

[0013] S5, traverse the partitioning mode in the partitioning list, return to S1 to enter the next depth of CU partitioning;

[0014] S6. End the current CU division process.

[0015] Furthermore, in the step S1, the encoding information of the current coding unit CU to be divided includes information including the division depth, the maximum depth allowed for division, the size and the channel;

[0016] Furthermore, in step S1, adding all feasible partitioning modes of the current CU to the partitioning list means that the partitioning mode of VVC includes quadtree QT, horizontal binary tree HBT, vertical binary tree VBT, horizontal ternary tree HTT and vertical ternary tree VTT partitioning. Among these 5 partitioning modes, the partitioning mode that the current CU can execute is selected and added to the partitioning list.

[0017] Furthermore, in step S2, the fast algorithm proposed by the present invention is only executed on the luminance CU with a size of 128×128 or 64×64, which can ensure that the coding efficiency will not be significantly reduced.

[0018] Furthermore, in step S3, the pixel difference value between the current CU and the best matching CU is calculated, which is specifically obtained by calculating the absolute error and SAD between the current CU and the best matching CU. The calculation formula of SAD is as follows:

[0019]

[0020] Where x and y are the horizontal and vertical coordinate values ​​of the pixel in the current CU respectively; W and H represent the width and height of the current CU respectively. If the current CU is a square, W is equal to H; f(x, y) is the brightness value at (x, y) in the current CU; g(x, y) is the brightness value at (x, y) in the best matching CU.

[0021] Furthermore, in step S3, if the pixel difference value is less than the threshold, that is, if the sum of absolute errors (SAD) is less than the threshold Th1, it means that a better prediction result can be obtained by predicting the current CU as a whole, and the division process of the current CU can be terminated in advance; the threshold Th1 is obtained through a large number of experiments, and the value is positively correlated with the encoding quantization parameter QP and the bit depth.

[0022] Further, in step S4, the texture difference of the sub-blocks in the horizontal direction and the texture difference of the sub-blocks in the vertical direction of the current CU are calculated by formula (2) to formula (4), and the current CU is divided into four sub-blocks from top to bottom in the horizontal direction, namely H1, H2, H3 and H4; the width of these four sub-blocks is equal to the width of the current CU, and the height of the four sub-blocks is equal to 1 / 4 of the height of the current CU; the current CU is divided into four sub-blocks from left to right in the vertical direction, namely V1, V2, V3 and V4, the width of these four sub-blocks is equal to 1 / 4 of the width of the current CU, and the height of the four sub-blocks is equal to the height of the current CU;

[0023]

[0024]

[0025]

[0026] In formula (2), DiffH 12 , DiffH 23 and DiffH 34 respectively represent the texture differences between adjacent sub - blocks H1 and H2, H2 and H3, H3 and H4 in the horizontal direction of the current CU; in formula (3), DiffV 12 , DiffV 23 and DiffV 34 respectively represent the texture differences between adjacent sub - blocks V1 and V2, V2 and V3, V3 and V4 in the vertical direction of the current CU; DiffH and DiffV obtained through formula (4) respectively describe the texture differences in the horizontal direction and vertical direction of the current CU.

[0027] Furthermore, in step S5, by comparing the magnitudes of the texture differences in two directions, it is determined whether to skip the multi - way tree partitioning process in a certain direction, and then the multi - way tree partitioning method in this direction is deleted from the partitioning list. Specifically, if DiffH < DiffV, skip the horizontal - direction multi - way tree partitioning process and delete the horizontal binary tree HBT and horizontal ternary tree HTT from the partitioning list; otherwise, skip the vertical - direction multi - way tree partitioning process and delete the vertical binary tree VBT and vertical ternary tree VTT partitioning methods.

[0028] A storage medium stores a computer program internally. It is characterized in that when the computer program is read by a processor, it executes the VVC fast CU partitioning method based on spatio - temporal characteristics as described in any one of the above.

[0029] The advantages and beneficial effects of the present invention are as follows:

[0030] Aiming at the problems of high computational complexity and long encoding time of the latest international video coding standard H.266 / VVC, the present invention proposes a CU fast division method based on space-time domain characteristics. First, the time domain correlation of the video is used to calculate the absolute error sum (SAD) between the current CU and its best matching CU obtained after motion estimation, and the motion attribute of the current CU is evaluated. The size of the SAD value is used to determine whether the current CU continues to be divided to the next depth, and all division processes of some CUs are terminated in advance; for the CU that needs to be divided to the next depth, the spatial domain characteristics of the video frame are used to analyze the texture differences of the sub-blocks in the horizontal and vertical directions of the current CU, determine its multi-tree division direction, and skip the multi-tree division process in the horizontal or vertical direction in advance. The present invention can be used for video coding under the VVC inter-frame coding configuration, and significantly reduce the VVC encoding time without almost reducing the encoding efficiency and encoding quality. It can be applied to application scenarios such as video conferencing and video surveillance that have high requirements for real-time encoding.

[0031] By limiting the use conditions of the fast algorithm of the present invention, the problem of excessive reduction in coding efficiency due to the use of the fast algorithm is avoided; the pixel difference between the current CU and the best matching CU obtained by motion estimation is calculated using SAD, the error generated by motion estimation is analyzed, and the division process is terminated in advance for the CU with a small pixel difference, so the CU division of the video frame can be significantly accelerated, which is particularly suitable for application scenarios such as video conferencing and video surveillance where the camera is fixed and has a large static background area. For CUs with more complex textures, the multi-tree division process in another direction is skipped by judging the main direction of the texture. For the determination of the main direction of the texture, only the corresponding pixel value subtraction method is used, so the additional amount of calculation required is very small, and the logic is simple and easy to implement. The CU fast division algorithm proposed in the existing relevant literature usually only uses the temporal correlation of the video, or only uses the spatial texture characteristics of the coded frame, and its acceleration effect needs to be improved. In terms of using the correlation between video frames, some literature uses the commonly used frame difference method to evaluate the difference between the current CU and the co-located CU of the reference frame, but ignores the motion properties of the video, resulting in a decrease in coding efficiency. The present invention makes full use of the spatial texture directional characteristics and temporal correlation of the video, and realizes fast CU division decision by combining the spatial and temporal characteristics. Therefore, the method in the present invention is highly innovative. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The present invention provides a flowchart of a VVC fast CU division method based on space-time domain characteristics in a preferred embodiment;

[0033] Figure 2 Sub-blocks obtained from different partitioning directions: (a) horizontal direction; (b) vertical direction. DETAILED DESCRIPTION

[0034] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0035] The technical solution of the present invention to solve the above technical problems is:

[0036] Figure 1 The present invention provides a VVC fast CU division method and storage medium based on space-time domain characteristics. The method of the present invention comprises the following steps:

[0037] S1. Obtain the coding information of the current coding unit CU to be divided in the coding frame, and determine whether the current CU can continue to be divided. If the current CU division depth is less than the set maximum division depth, continue to divide, add all feasible division modes of the current CU to the division list, and enter step S2; otherwise, enter step S6;

[0038] S2, determine the size and type of the current CU, if the current CU is a 128×128 or 64×64 brightness CU, go to step S3, otherwise, go to step S5;

[0039] S3, read the pixel value of the best matching CU obtained by motion estimation of the current CU in the reference frame, calculate the pixel difference value between the current CU and the best matching CU, if the pixel difference value is less than the threshold, go to step S6, otherwise go to step S4;

[0040] S4, respectively calculating the texture difference of the sub-block in the horizontal direction and the texture difference of the sub-block in the vertical direction of the current CU, and judging whether to skip the multi-tree partitioning process in a certain direction by comparing the sizes of the texture differences in the two directions, and then deleting the multi-tree partitioning mode in the partitioning list;

[0041] S5, traverse the partitioning mode in the partitioning list, return to S1 to enter the next depth of CU partitioning;

[0042] S6. End the current CU division process.

[0043] Preferably, in step S1, the encoding information of the current coding unit (CU) to be divided includes information such as division depth, maximum depth allowed for division, size and channel;

[0044] Preferably, in step S1, all feasible partitioning modes of the current CU are added to the partitioning list, which means that the partitioning modes of VVC include quadtree (QT), horizontal binary tree (HBT), vertical binary tree (VBT), horizontal ternary tree (HTT) and vertical ternary tree (VTT) partitioning, and a partitioning mode that can be executed by the current CU is selected from these five partitioning modes and added to the partitioning list;

[0045] Preferably, in step S2, in order to balance the coding efficiency and acceleration effect of the video encoder, the present invention only performs fast partitioning on the luminance CU with a size of 128×128 or 64×64;

[0046] Preferably, in step S3, the pixel difference value between the current CU and the best matching CU is calculated by calculating the sum of absolute differences (SAD) between the current CU and the best matching CU. The calculation formula of SAD is as follows:

[0047]

[0048] Among them, x and y are the horizontal and vertical coordinate values ​​of the pixel point in the current CU respectively; W and H represent the width and height of the current CU. If the current CU is a square, W is equal to H; f(x,y) is the brightness value at (x,y) in the current CU; g(x,y) is the brightness value at (x,y) in the best matching CU.

[0049] Preferably, in step S3, if the pixel difference value is less than the threshold, that is, if the sum of absolute errors (SAD) is less than the threshold Th1, it means that a better prediction result can be obtained by predicting the current CU as a whole, and the current CU division process can be terminated in advance. Th1 can be obtained through a large number of experiments, and the value is positively correlated with the coding quantization parameter (QP) and the bit depth.

[0050] Preferably, in step S4, the texture difference of the sub-blocks in the horizontal direction and the texture difference of the sub-blocks in the vertical direction of the current CU are calculated by formula (2)-formula (4). The current CU is divided into four sub-blocks from top to bottom in the horizontal direction, namely H1, H2, H3 and H4. The width of these four sub-blocks is equal to the width of the current CU, and the height of the four sub-blocks is equal to 1 / 4 of the height of the current CU; the current CU is divided into four sub-blocks from left to right in the vertical direction, namely V1, V2, V3 and V4, and the width of these four sub-blocks is equal to 1 / 4 of the width of the current CU, and the height of the four sub-blocks is equal to the height of the current CU.

[0051]

[0052]

[0053]

[0054] In formula (2), DiffH 12 、DiffH 23 and DiffH 34 Respectively represent the texture differences between adjacent sub-blocks H1 and H2, H2 and H3, H3 and H4 in the horizontal direction of the current CU; in formula (3), DiffV12 , DiffV 23 and DiffV 34 respectively represent the texture differences between adjacent sub-blocks V1 and V2, V2 and V3, V3 and V4 in the vertical direction of the current CU. The DiffH and DiffV obtained through formula (4) respectively describe the texture differences in the horizontal direction and the vertical direction of the current CU.

[0055] Preferably, in step S4, by comparing the magnitudes of the texture differences in two directions, it is determined whether to skip the multi-way tree partitioning process in a certain direction, and then the multi-way tree partitioning method in this direction is deleted from the partitioning list. Specifically, if DiffH < DiffV, skip the horizontal multi-way tree partitioning process and delete the horizontal binary tree (HBT) and horizontal ternary tree (HTT) from the partitioning list; otherwise, skip the vertical multi-way tree partitioning process and delete the vertical binary tree (VBT) and vertical ternary tree (VTT) partitioning methods from the partitioning list.

[0056] A storage medium stores a computer program internally. When the computer program is read by a processor, it executes the method of any one of the above.

[0057] Computer-readable media includes both permanent and non-permanent, removable and non-removable media and can be implemented by any method or technology for information storage. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0058] It should also be noted that the term "comprises", "comprising", or any other variation thereof is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but also other elements not expressly listed, or elements that are inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0059] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A fast CU partitioning method for VVC based on spatio-temporal characteristics, characterized in that, it includes the following steps: S1. Obtain the coding information of the current coding unit CU to be partitioned in the coded frame, and determine whether the current CU can be further partitioned. If the partitioning depth of the current CU is less than the set maximum partitioning depth, continue partitioning, add all feasible partitioning modes of the current CU to the partitioning list, and enter step S2; otherwise, enter step S6; S2. Determine the size and type of the current CU. If the current CU is a luma CU of 128×128 or 64×64, enter step S3; otherwise, enter step S5; S3. Read the pixel values of the best matching CU obtained by motion estimation of the current CU in the reference frame, calculate the pixel difference value between the current CU and the best matching CU. If the pixel difference value is less than the threshold, enter step S6; otherwise, enter step S4; S4. Calculate the texture differences of the sub-blocks in the horizontal direction and the vertical direction of the current CU respectively. By comparing the magnitudes of the texture differences in these two directions, determine whether to skip the multi-tree partitioning process in a certain direction, and then delete the multi-tree partitioning method in that direction from the partitioning list; S5. Traverse the partitioning modes in the partitioning list, return to S1 and enter the CU partitioning at the next depth; S6. End the partitioning process of the current CU; In step S4, the texture difference of the sub-blocks in the horizontal direction and the texture difference of the sub-blocks in the vertical direction of the current CU are calculated by formula (2) to formula (4). The current CU is divided into four sub-blocks from top to bottom in the horizontal direction, namely H 1 , H 2 , H 3 and H 4 ; The width of these four sub-blocks is equal to the width of the current CU, and the height of the four sub-blocks is equal to 1 / 4 of the height of the current CU; the current CU is divided into four sub-blocks from left to right in the vertical direction, namely V 1 、V 2 、V 3 and V 4 , the width of these four sub-blocks is equal to 1 / 4 of the current CU width, and the height of the four sub-blocks is equal to the height of the current CU; In formula (2), DiffH 12 、DiffH 23 and DiffH 34 Respectively represent the adjacent sub-blocks H in the horizontal direction of the current CU 1 With H 2 , H 2 With H 3 , H 3 With H 4 The texture difference between 12 、DiffV 23 and DiffV 34 Respectively represent the adjacent sub-blocks V in the vertical direction of the current CU 1 With V 2 、V 2 With V 3 、V 3 With V 4 The texture difference between the two CUs; ​​DiffH and DiffV obtained by formula (4) respectively describe the texture difference in the horizontal direction and the texture difference in the vertical direction of the current CU; In step S5, by comparing the magnitudes of the texture differences in two directions, determine to skip the multi-tree partitioning process in a certain direction, and then delete the multi-tree partitioning method in that direction from the partitioning list. Specifically: If DiffH < DiffV, skip the multi-tree partitioning process in the horizontal direction, and delete the horizontal binary tree HBT and the horizontal ternary tree HTT from the partitioning list; otherwise, skip the multi-tree partitioning process in the vertical direction, and delete the vertical binary tree VBT and the vertical ternary tree VTT partitioning methods from the partitioning list.

2. A fast CU partitioning method for VVC based on spatio-temporal characteristics according to claim 1, characterized in that, in step S1, the coding information of the current coding unit CU to be partitioned includes information such as the partitioning depth, the maximum allowable partitioning depth, the size, and the channel.

3. A fast CU partitioning method for VVC based on spatio-temporal characteristics according to claim 1 or 2, characterized in that, in step S1, adding all feasible partitioning modes of the current CU to the partitioning list means: The partitioning modes of VVC include quadtree QT, horizontal binary tree HBT, vertical binary tree VBT, horizontal ternary tree HTT, and vertical ternary tree VTT partitioning. Select the partitioning modes that the current CU can execute from these 5 partitioning modes and add them to the partitioning list.

4. A fast CU partitioning method for VVC based on spatio-temporal characteristics according to claim 1, characterized in that, in step S3, calculating the pixel difference value between the current CU and the best matching CU is specifically obtained by calculating the sum of absolute differences SAD between the current CU and its best matching CU. The calculation formula of SAD is as follows: Where x and y are the horizontal and vertical coordinate values ​​of the pixel in the current CU respectively; W and H represent the width and height of the current CU respectively. If the current CU is a square, W is equal to H; f(x, y) is the brightness value at (x, y) in the current CU; g(x, y) is the brightness value at (x, y) in the best matching CU.

5. The VVC fast CU division method based on space-time characteristics according to claim 4, It is characterized in that In step S3, if the pixel difference value is less than the threshold, it means if the sum of absolute differences (SAD) is less than the threshold Th 1 , indicating that a better prediction result can be obtained by predicting the current CU as a whole. At this time, the division process of the current CU can be terminated in advance; the threshold Th 1 Through a large number of experiments, it is found that this value is positively correlated with the encoding quantization parameter QP and the bit depth.

6. A storage medium storing a computer program, It is characterized in that When the computer program is read by a processor, the VVC fast CU division method based on space-time domain characteristics of any one of claims 1 to 5 is executed.