Research On Fast Algorithm For High Efficiency Video Coding

High Efficiency Video Coding(HEVC) is a new generation of video coding standard.It is developed by the Joint Collaborative Team on Video Coding(JCT-VC)with the aim to meet the demand for higher quality video.Compared to the previous generation video coding standard H.264/AVC,HEVC employs a lot of new coding technologies,for example,quad tree-based recursive partitioning of Coding Unit(CU),which make it has higher compression ratio.However, it also leads to higher encoding complexity.Hence,the research of fast encoding algorithm is crucial to HEVC.Firstly, a brief introduction to the framework and the key techniques of the HEVC encoding is maded in the paper. After that,a series of fast algorithms are presented including intra encoding and inter encoding according to the encoding features of HEVC.The coding optimization of intra prediction covers two aspects: macroscopic optimization and microcosmic optimization.In the macroscopic level: first,in order to reduce depth range of current Largest Coding Unit(LCU),we analyze the segmentation structure similarities between the current LCU and spatial neighborhood LCUs;second,it is based on Bayes decision theory that the early CU splitting termination are performed at each CU depth level by analyzing the relationship between the Sum of Absolute Transformed Difference(SATD) and the CU splitting.Experimental results show that,compared with the original algorithm of HM10.1, the proposed fast intra macroscopic coding algorithm can save an average of 34.7% coding time while the BDBR increases about 1.00%,and the Y BDPSNR decreases about 0.04 dB under the configuration of All Intra(AI)encoding.In the microcosmic level: first,according to the correlation between the texture of the Prediction Unit(PU) and mode cost,as well as the results of the statistical analysis,the fast mode decision algorithms of the Rough Mode Decision(RMD) and the Rate Distortion Optimization(RDO) are proposed to decrease the number of thecalculative candidate modes;then,we propose an optimization algorithm to early termination the Transform Unit(TU) splitting basing on the correlation among texture of TU and TU splitting.The experimental results show that,under AI configuration,compared with the original algorithm of HM10.1, the proposed fast intra microcosmic coding algorithm, can achieve a reduction of 31.9% encoding time on average while incurring 1.55% increment in BDBR and 0.08 dB loss in Y BDPSNR.In order to reduce the encoding time more efficiently,a novel fast intra prediction algorithm is suggested by combining the proposed fast macroscopic algorithm and the microcosmic algorithm.The experimental results show that,under AI configuration,the proposed fast intra coding algorithm can cut encoding time down on 55.3% on average with maintain the same rate distortion performances as the original algorithm of HM10.1.In the inter prediction coding optimization level:first,it is based on the strong correlations of the skip mode and the texture of CU, as well as the information of motion vector(MV) which is spatially adjacent to the CUs that we can terminate CU splitting in advance;second, analyze the kinds of PU mode rate distortion cost(RDcost)distribution,adaptive threshold inter-mode decision strategies are proposed including early skip mode decision and early Asymmetric Motion Partition(AMP) mode skip.Experimental results demonstrate that,under the inter coding cases,compared with the original algorithm of HM10.1,the proposed fast inter coding algorithm can reduce the encoding time by 42.2% with only 0.68% increment in BDBR,and 0.02 dB loss in Y BDPSNR on average.Finally,the new achievements of the whole thesis and the prospect of the next plan of the research are summarized.