Block Partition Fast Algorithm For HEVC

High Efficiency Video Coding (HEVC) is currently being prepared as the most advanced and most efficient video coding standard. As a new generation of video coding standard, the main goal of the HEVC standardization effort is to enable significantly improved compression performance relative to H.264/AVC—in the range of 50% bit-rate reduction for equal perceptual video quality. HEVC adopts many new coding technologies in which a quadtree-based coding block partitioning structure introduced three basic units that is Coding Unit(CU), Prediction Unit(PU) and Transform Unit(TU). This new technique can improve the coding efficiency significantly for it is more flexible to adapt to various texture characteristics of images and can obtain more accurate division in HEVC. However, it also introduces a great computation complexity. This paper proposed fast algorithms for block partition selection of HEVC in intra and inter prediction, as follows:1、For intra prediction, this paper proposes a fast CU splitting algorithm which can narrow CU depth range and early terminate the CU splitting. This algorithm narrows CU depth range by judging the complexity of texture characteristics of images based on Sobel operator, then if the RD cost of current depth CU is less than the corresponding RD cost threshold in the CU depth range, it will early terminate the splitting of CU or PU. This algorithm is able to save about 59% encoding time on average with only 1.2% increasing on bit-rate and almost the same value of PSNR which can observably improve the coding efficiency.2、In inter prediction, the recursive partitioning process is more complex than intra prediction for it obtains best mode by the comparison of merge mode, inter mode and intra mode. For inter prediction, this paper proposes three fast algorithms, as follows:(1) It proposes a fast CU splitting algorithm which can early terminate the CU division based on the judgement of RD cost. That is to say that if the RD cost of current depth CU is less than the corresponding RD cost threshold, there is no need to split CU in more larger depth;(2) It proposes a two-stage fast algorithm to early terminate CU encoding. If the best prediction mode of current depth CU is SKIP mode, it will early terminate the CU splitting. Otherwise, it will judge whether the RD cost of current depth CU is less than the corresponding RD cost threshold for early termination of the CU splitting in a second stage;(3) It proposes a fast algorithm to early terminate PU mode encoding in current CU depth, by detecting the information of Coded Block Flag(CBF) of PU and the RD cost of PU mode of inter prediction in current CU depth;This paper proposes CU splitting decision fast algorithms according to the relevant information of the splitting of CU and PU in intra and inter prediction. As a result, the algorithms can reduce the complexity introduced by CU recursive partitioning in HM and save encoding time efficiently.