Research On Low-complexity Encoding Techniques For The New Video Coding Standard HEVC

Video has become the main carrier of information for communication and reception in daily life. The amount of video information data is very large, in order to reduce the time consumption for transmission and to realize the real-time transmission of information, the video coding technique which efficiently compress video signal has flourished. The video coding standard H.264/AVC, which is widely used in the first ten years of 21 st century, cannot satisfy the compression for HDTV and UHDTV, therefore, a new generation of international video coding standard H.265/HEVC(High Efficiency Video Coding) arises at the historic moment. HEVC adopts many advanced techniques to realize the optimization on compression performance, compared with H.264/AVC, it doubles the compression efficiency. However, HEVC has a high computational complexity and it requires an unbearable encoding time. In this dissertation, two improved algorithms are proposed for HEVC Intra prediction coding, one improved algorithm is proposed for HEVC Inter prediction coding. The improved algorithms can reduce the computational complexity of HEVC and improve the coding speed. The main content of this dissertation is as follows:1. A fast Coding Unit(CU) partition algorithm for HEVC Intra prediction coding.In this dissertation, we use Alternating Current(AC) energy to measure the texture complexity of Largest Coding Unit(LCU), and remove the least possible partition modes. Besides, we reduce the coding depth range by spatial correlation of coding depth. We can skip the least possible partition modes and terminate coding early if needed. Compared with HM-13.0, the proposed algorithm achieves 23.40% reduction in coding time with negligible degradation in terms of coding efficiency(BDBR increases 0.42%, BDPSNR reduces 0.02 d B).2. A fast mode decision algorithm for HEVC Intra prediction coding.In this dissertation, we construct a candidate mode list for the current Prediction Unit(PU) by prediction mode correlation of different PUs in adjacent coding layers and the temporal correlation of prediction modes, and reduce the number of prediction modes before Rough Mode Decision(RMD). Besides, we remove some modes before Residual Quad-tree Transform(RQT) by comparing Hadamard cost. Compared with HM-13.0, the proposed algorithm achieves 38.98% reduction in coding time with negligible degradation in terms of coding efficiency(BDBR increases 1.51%, BDPSNR reduces 0.09 d B). Besides, the fast CU partition algorithm and the fast mode decision algorithm are combined. Compared with HM-13.0, the combined fast algorithm achieves 57.79% reduction in coding time with negligible degradation in terms of coding efficiency(BDBR increases 1.81%, BDPSNR reduces 0.10 d B).3. A fast PU partition algorithm for HEVC Inter prediction codingIn this dissertation, we reduce the number of PU partition modes which need Rate Distortion Cost(RDC) computation for the current CU by using the temporal correlation between the PU partition modes. When the patition mode of the co-located CU is 2N×2N, a more detailed prediction for the PU partition modes of current CU is performed by considering the resolution of the video image, Motion Vector(MV) of the current CU and the coding depth. In this way, we can avoid massive RDC computation of the least possible PU partition modes. Compared with HM-13.0, when we adopt the Lowdelay configuration, the proposed algorithm achieves 45.55% reduction in coding time with negligible degradation in terms of coding efficiency(BDBR increases 1.89%, BDPSNR reduces 0.06 d B), and when we adopt the Randomaccess configuration, the proposed algorithm achieves 43.80% reduction in coding time with negligible degradation in terms of coding efficiency(BDBR increases 1.51%, BDPSNR reduces 0.05 d B).