Research Of Fast Encoding Algorithm Based On HEVC

In recent years, digital video, especially digital high-definition video has become a leading format of media content in many areas of consumption. In order to facilitate transmission of digital video, video coding is one of the key technology. HEVC as the next new generation standard, was jointly developed by the two video coding standard makers which were ITU-T and ISO in January 2010. HEVC still retains the traditional framework of video coding, and at the same time adopts a lot of new techniques research on video compression in recent years, which makes the encoding performance reach 2 times than the H.264/AVC standard. However, the improvement of video compression performance has brought huge encoding complexity. Therefore, how to reduce the computational complexity under the premise of not reducing the HEVC intra coding performance, and save the encoding time, which has become an important research topic in HEVC video coding.This paper firstly has carried on the research of HEVC video coding, and introduced respectively the coding structure of HEVC and its key technology adopted which including quad-tree structure, intra prediction, inter prediction, transform and quantization, entropy coding and the loop filter. Then it mainly lucubrates part of intra prediction of HEVC video coding, including the analysis of distortion optimization process of detailed analysis and intra mode selection flow rate.On this basis, aim at the high complexity of HEVC intra prediction, we propose two kinds of HEVC intra prediction algorithm, which greatly speeds up the coding speed and improves the coding efficiency. Firstly, using the CU depth range of video sequence and the correlation of the adjacent encoding blocks coding forecast the current CU depth, and adaptive decide to early terminate CU partition or skip some unnecessary division of CU. In addition, TU depth range is affected by the probability of each TU depth, and a redundant process removed in the TU selection process. Experiments show that the proposed algorithm in this paper can save about 14% of the encoding time on average at the expense of 0.75% bitrate increased and 0.041dB PSNR decreased on average than the original algorithm. Secondly, by using the coefficients of integer transform to determine the texture direction of prediction block, we propose a new algorithm which using integer transform for fast implementation intra prediction in HEVC. It filtrates prediction modes using the feature of transform domains, and reduces the number of candidate mode set according to the correlation of prediction mode among the adjacent block, and finally carries on the RDO mode selection process. Experiment show that compared with the original algorithm, the proposed algorithm in this paper provides about 20% reduction of intra coding time with neglecting degradation of coding performance. Finally, we integrate the above two schemes into a full fast intra prediction method. The experimental results show that such integrated fast algorithm saves averagely about 33% of encoding time with negligible bitrate increase and PSNR loss, which greatly speeds up the encoding speed and effectively improves the coding efficiency.