Research On Rate-distortion Optimization Techniques For The H.265/HEVC Video Coding Standard

Rapid developments of information and multimedia technologies have witnessed an explosion of various video services,especially for high-definition video applications,which further bring great challenges to the current video coding standards.Therefore,the ITU-T Video Coding Experts Group(VCEG)and the ISO/IEC Moving Picture Expert Group(MPEG)standardization organizations working together as a partnership and jointly formulated the newest video coding stand H.265/HEVC.Compared with the previous video coding standard H.264/AVC,it can achieve a bitrate reduction of nearly 50% by utilizing more flexible coding structures and lots of advanced coding techniques.However,it also requires much more computational complexity compared with the previous standards.This dissertation mainly focuses on the rate-distortion optimization techniques and the rate control module in the H.265/HEVC encoding process,and proposes a high-speed implementation algorithm of rate-distortion optimized quantization(RDOQ)for H.265/HEVC,an adaptive quantization parameter(QP)allocation method for H.265/HEVC based on inter-frame dependency and an efficient frame-level bit allocation algorithm for H.265/HEVC.The major contributions of this dissertation are summarized as follows:1.Rate-distortion optimized quantization(RDOQ)is generally utilized in video coding for achieving higher coding efficiency which requires considerable complexity in practice to calculate rate-distortion(RD)costs for multiple candidates of quantized levels.This dissertation proposes a fast RDOQ algorithm with low computational complexity for H.265/HEVC.The candidates of quantized levels that are unlikely to be selected are firstly excluded.Then,instead of calculating the RD costs for two candidates of quantized levels separately,their RD cost differences are derived.A bit-rate estimation method is then used to accurately calculate the length of coding bits in the RD cost function for avoiding the high-computation-cost process of the context-adaptive binary arithmetic coding(CABAC).Experimental results show that,with a negligible degradation of coding efficiency,the proposed algorithm is faster than the original RDOQ by 74.7% in average.2.Rate-distortion optimization(RDO)is widely applied in video coding process which aims at minimizing the coding distortion at a target bitrate.Conventionally,the RDO scheme is simply performed on each frame individually and independently for avoiding high computational complexity.However,due to an extensive use of temporal/spatial predictions,strong coding dependencies are existed among neighboring frames.Therefore,a global RDO problem with inter-frame dependency is required for achieving higher coding performance.In this dissertation,an adaptive frame-level quantization parameter(QP)selection algorithm is proposed for H.265/HEVC by taking into account the inter-frame dependency.A model considering the energy of prediction residuals is established for measuring the inter-frame dependency.Based on which the dependent RDO scheme is then analyzed for the hierarchical coding structure in H.265/HEVC according to its referencing structure.Finally,the QP and the corresponding Lagrangian multiplier for each coding frame are determined considering the total impact of its coding distortion on that of the future frames in the encoding order.Experimental results show that the proposed algorithm can achieve an average BD-rate of-3.49% over HM with a negligible encoding time increment.Furthermore,the proposed algorithm can also reduce the quality fluctuation of the coded video.3.Rate control plays a critical role in video coding for regulating the coded bit-stream to achieve the target bitrate,which mainly involves two basic steps: bit allocation and QP calculation,and the former is a key element to directly affect the following operations and performance of rate control.In this dissertation,an efficient frame-level bit allocation algorithm is proposed for H.265/HEVC.Firstly,by analyzing the optimization problem of frame-level bit allocation,two factors have been found to be related to the frame-level bit allocation,which are inter-frame dependency and target bitrate.Then,a measure of inter-frame dependency considering the intra-frame and inter-frame complexities is put forward to reflect the correlation among consecutive frames.Finally,according to the results of allocation weights through the dynamic programming method,the relationships among the allocation weights and the two factors are investigated,and based on which the efficient bit allocation model is established.Experimental results show that the proposed frame-level bit allocation algorithm can achieve an average BD-rate of-5.23% over the original bit allocation algorithm in HM with a negligible encoding time increment.Furthermore,the proposed scheme can also reduce the GOP-level bitrate fluctuation of the coded video.