Research On Fast Algorithm For VVC Intra Coding

With the development of the video market,some video contents including 4K/8K Ultra High Definition(UHD)video,Virtual Reality(VR),High Dynamic Range(HDR)video and 360-degree video have become more and more popular.However,it has also led to a sharp increasing in the bandwidth required for video transmission and the capacity required for video storage.Video coding technologies are facing great challenges.In order to meet the needs of the video market,the International Telecommunication Union and the International Organization for Standardization jointly conducted research on the next-generation video coding standards,and launched the Versatile Video Coding(VVC)standard in the summer of 2020.Compared with the previous generation of High Efficiency Video Coding(HEVC)standards,VVC reduce the bit rate by 50% without affecting the video quality.Achieving such a large increasing in coding efficiency stems from the adoption of a series of new coding techniques by VVC,but the new coding techniques have also caused a sharp increasing in coding complexity.According to the report,under the All Intra(AI)configuration,the coding complexity of VVC is 10-20 times that of HEVC.In order to significantly reduce the complexity of VVC intra coding without affecting the coding efficiency and image quality,this thesis studies the coding unit(CU)partition and intra prediction technologies,and proposes corresponding fast algorithms.The specific works are as follows :1.For CU partition,this thesis proposes a fast CU partition decision(FCPD)algorithm.First,analyzing the technical principle and implementation process of CU partition.Then,performing statistics on the relevant information in the implementation process,and analyze the factors that affect CU partition.Next,selecting representative features based on the analysis results,and designing corresponding acceleration strategies.Finally,training a random forest(RF)classifier to predict the partition mode of CU.The experimental results show that the algorithm effectively reduces the coding complexity under the condition of ensuring the coding efficiency.Specifically,this algorithm can save about 57% of coding time,BD-BR only increases by 1.21%,and BD-PSNR only decreases by 0.12 d B.2.For intra prediction,this thesis proposes a fast intra mode decision(FIMD)algorithm.First,analyzing the principle of intra prediction technology and the implementation process.Then,performing statistics on the relevant information in the implementation process,and analyzing the factors that affect the selection of the intra prediction mode according to the statistical results.Finally,designing the corresponding acceleration strategy according to the analysis results to optimize the search range and search method.The experimental results show that the algorithm effectively reduces the coding complexity under the condition of ensuring the coding efficiency.Specifically,this algorithm can save about 25% of coding time,BD-BR only increases by 0.6%,and BD-PSNR only decreases by 0.03dB.