Research On Complexity Control Method Of High Definition Video Coding

The continuous development of multimedia technology can provide a clearer,smoother and richer visual experience,which also leads to the amount of video data increase dramatically.In the case of limited transmission bandwidth,the previous video coding standards cannot meet the existing requirements.Consequently,High Efficiency Video Coding(HEVC)has been developed.HEVC significantly improves the compression efficiency by a large number of new coding techniques over the preceding H.264 video coding standard,but at the cost of extremely huge computational complexity.Currently,the real-time application of high-definition video has been greatly hindered due to the unstable computing power of power-limited devices.Thus,the computational complexity control of HEVC is an urgent problem to be solved.In this paper,two high-definition video coding complexity control schemes are proposed.(1)A hierarchical complexity control algorithm for HEVC based on coding unit(CU)depth decision is proposed.Firstly,the coding complexity of group of pictures layer and the frame layer is allocated and controlled according to the target complexity and the constantly updated reference coding time.Secondly,Secondly,the maximal depth is assigned to the coding tree unit(CTU)on the basis of the correlation between the residual information and the optimal depth by establishing the complexity-depth model.Then,a CU smoothness decision and an adaptive low bit threshold decision are proposed for early terminating the unnecessary traversal process.Moreover,an adaptive upper bit threshold decision is designed to protect the CU decision of important CUs for maintaining the visual quality.The experimental results show that this algorithm can control the encoding complexity between 50% and 100% of the original encoding complexity with high control accuracy and the negligible rate distortion performance loss.Meanwhile,this algorithm outperforms state-of-the-art algorithms in terms of control accuracy and encoding quality.(2)A complexity control method of HEVC based on random forest is proposed.Due to the diversity of video content,it is difficult to accurately predict the optimal depth and mode of each CU through a model.In this paper,two classes of random forest classifiers with three kinds of prediction accuracy are trained respectively to provide multiple coding configurations for encoding each CTU.And then the average depth-complexity model is established to allocate complexity to each CTU.The coding configuration of each CTU is determined by flatness,average depth,bit and CTU level cumulative coding error,and used to complete the reduce of complexity.Experimental results show that the algorithm can control the coding complexity to 40% of the original encoder.In addition,compared with the state-of-the-art algorithms,the proposed algorithm has better rate-distortion performance.