Research On Coding Block Partition Algorithm Based On CU Characteristics For VVC

Versatile Video Coding(VVC)has been officially released in July 2020.It not only significantly improves the coding efficiency,but also remarkablely increases the computational complexity.Related work shows that the multi-type Tree(MTT)partition structure introduced by VVC takes up a great deal of coding time.Compared with the quadtree partition structure,MTT increases the coding time by about 9 times.In order to reduce the computational complexity of coding block partitioning in VVC,based on the characteristics of Coding Unit(CU),the process of multi-type tree partitioning and Intra sub-partition(ISP)Coding mode decision is studied in this paper,and two fast decision algorithms are designed.The main content and innovations are as follows:(1)To solve the problem of the large computational complexity in MTT(multi-type tree)partitioning,the relationship between the coding block partitioning and pixel distribution has been studied.It is found in this paper that the application of binary tree and ternary tree in horizontal or vertical direction is closely related to the pixel difference of the sub-block of CU.Based on this observation,a fast multi-type tree decision algorithm based on sub-block pixel difference is proposed,and a new metric named the pixel difference of sub-blocks(SBPD)is defined to measure the sub-block pixel difference of CU in different splitting modes.By comparing the SBPD of CU in horizontal and vertical split for binary tree and ternary tree respectively,the selection of binary tree and ternary tree is decided in advance,so as to skip some redundant splitting modes and save coding time.In addition,the statistical method is used to optimize the decision threshold.Experimental results show that compared with the reference model VTM-8.0,the proposed algorithm can save 27% coding time on average,and the BD-rate is only increased by 0.55%.(2)ISP coding mode is a new tool introced in VVC.In this paper,by analyzing the rate-distortion cost of CU with different texture characteristics in ISP coding mode,it is found that the coding performance of CU with simple texture is not significantly improved in ISP coding mode.Based on this observation,a fast decision algorithm for ISP coding mode based on the CU texture complexity is proposed.By comparing the texture complexity of current CU,it can be decided in advance whether the ISP coding mode is applied to this CU or not,so as to reduce the frequency of using ISP coding mode and save coding time without significantly reducing the performance.A method for calculating the CU texture complexity based on interval sampling is designed in the algorithm,either,and the decision threshold is optimized by using the rate distortion cost.Experimental results show that compared with the reference model VTM-8.0 the proposed algorithm can save 7% coding time on average,and the BD-rate is only increased by 0.09%.