Fast Depth Selection Algorithm Based On Hybrid Temporal And Spatial Domain For HEVC

Today, with the rapid development of information technology and multimedia video, the requirement of video quality has increased sharply. The previous video coding standard is not able to meet current needs any more. In order to fulfill the challenges i n the field of video coding, ISO and ITU released a new generation of video coding standard-HEVC in 2013. While retaining the basic framework of the previous generation of video coding standard, HEVC succeeded to make further improvements on many aspects of intra prediction, inter prediction and so on. HEVC can increase 50 percent in rate saving, while guarantees the same reconstruction quality. HEVC has obvious performance advantages especially dealing with ultra-high definition video, high-definition video and home theater.HEVC paid a huge cost in coding complexity even though it ensures a certain reconstruction quality and encoding performance. On the basis of same resolution, HEVC costs twice more time than H.264/AVC. So, how to reduce the time consumi ng of HEVC has become the hotspot of HEVC research area. This paper focuses on the splitting strategy of coding units, and analyzes the similarity of split depth between the current CU and the same location CU in adjacent block. Two fast depth determine algorithms of inter prediction are proposed as followed:Firstly, a CU depth quick selection algorithm based on temporal correlation is proposed. This algorithm can determine the depth of current CU quickly according to the depth of current col-CU and its surrounding depth to reduce the number of CU splitting.Secondly, a depth fast determine algorithm based on hybrid temporal and spatial domain algorithm is proposed. Based on the analysis of spatial feature of motion vector in inter prediction and the similarity of LCU nearby area, it is obvious that the texture similarity between adjacent frames is very strong. We determine the depth range of LCU, to reduce the number of LCU depth splitting. When current region has weak texture similarity, the fast depth algorithm based on CU level is in use, to further reduce coding time and reduce coding complexity.To verify the effectiveness of the above algorithm, we test different standard video sequences to conduct extensive testing in different QP. The experimental results show that the two fast depth determine algorithms can achieve same video quality while reducing computing time greatly. The first algorithm reduce the average PSNR of 0.08 d B, the PSNR of second algorithm declined about 0.09 d B. The coding complexity decreased 52.18 percent and 54.08 percent respectively. The experimental results show the proposed algorithm for inter prediction optimization can reduce the coding complexity of HEVC, also has a positive meaning for the applicability of HEVC.