Research On Computing Complexity Control Algorithm For H.265/HEVC

In order to meet the requirements of the high-definition video coding,H.265/HEVC,a high-efficiency video coding standard,improves coding efficiency by introducing many new coding techniques.However,these coding techniques inevitably increase the computing complexity of the coding as well.In practice,different applications have different requirements in encoding computing complexity.For example,a mobile application requires low computing complexity for encoding,while the video player in medical device pays more attention to the encoding quality.To make a trade-off between the coding performance and the computing complexity according to the different application requirements,the computing complexity control scheme for H.265/HEVC is proposed in this paper.The main contributions in this paper are presented as follows:(1)Based on the error propagation relationships and the hierarchical prediction structure in H.265/HEVC,this paper proposes two improved fast motion estimation algorithms,which makes the foundation for overall computing complexity control scheme.Firstly,we analyze the error propagation relationships between the Coding Units(CUs),and then propose an error propagation resilience algorithm which eliminates the rate-distortion loss due to error propagation.Besides,a hierarchical range adjustment algorithm that adjusts the values of motion estimation between hierarchical layers is also proposed.Simulation results show that our methods greatly improve the rate-distortion performances of traditional motion estimation works whilst keeping almost intact complexity reductions.(2)A three-level non-uniform computing complexity allocation scheme for multi-view video is proposed,which allocates different complexity to different views,frames and coding units.The traditional average complexity allocation scheme,which does not allocate complexity according to the hierarchical prediction structure of inter-view and inter-layer,results in the relatively important video views,frames and coding units having the relatively low complexity,thereby causing the unnecessary loss of encoding quality.To address this problem,this paper presents a non-uniform allocation scheme for computational complexity.Firstly,according to the hierarchical prediction structure between inter-view and intra-view,this paper constructs the rate-distortion dependency model for multi-view video,which is utilized to achieve non-uniform complexity allocation of view level and frame level.Secondly,based on the prediction uncertainty relationship of coding units in different regions within the video frame,the non-uniform complexity allocation of coding unit level is achieved.(3)By combining with the above algorithms,this paper proposes an overall computing complexity control scheme for multi-view video,which includes complexity allocation and fast algorithms.The former is the three-level non-uniform complexity allocation scheme in(2).Besides the fast motion estimation algorithm in(1),by using the imulti-view video coding mode selection principle,the latter also proposes a multi-view video mode selection fast algorithm based on mode mapping,which includs coding unit optimal mode prediction algorithm,the mapping algorithm based on cost function and mode selection candidate set determination algorithm,to simplify the complicated mode selection process.To verify the proposed schemes,we implement them on recent H.265/HEVC standard reference software.The results show that the above schemes can achieve good computing complexity control performances and have high practical value.