Research On Fast Algorithm For Depth Video Coding In 3D-HEVC

HEVC,as a new generation of video coding standard,improves nearly 50%coding efficiency comparing to the previous H.264/MPEG-4 AVC standard.3D-HEVC,as the 3D video extension of HEVC,inherits the coding techniques of HEVC and meanwhile introduces inter-view and inter-component coding tools for efficient coding of dependent views and depth data,respectively.However,the improvement of coding efficiency results in the increase of computational complexity,which brings a certain affection on the application of 3D-HEVC standard.It is of vital importance to research fast and efficient coding algorithms of 3D-HEVC.This paper introduces the background and history of video compression coding and elaborates the coding frameworks and key technologies of HEVC and 3D-HEVC standards.Based on the deep analysis of the key technologies of 3D-HEVC and the video characteristics of MVD,the researches are mainly foucses on the computational complexity reduction of depth video coding.This paper studies on the fast algorithm of depth video coding in 3D-HEVC from two directions: inter prediction and intra prediction.Depth video has different characteristics from texture video,which contains a lot of smooth regions and sharp boundaries.Multiview video has inter-view correlation.This paper deeply analyzes the inter-frames grayscale similarity and inter-view correlation for depth video,and proposes a fast algorithm for depth video coding based on grayscale similarity and inter-view correlation.First,the range of CU recursive depth level and PU mode decision process can be determined based on the grayscale similarity between inter frames.Moreover,the grayscale similarity and inter-view correlation are jointly used for depth map coding of dependent views to early decide the best prediction unit(PU)mode.At last,in order to guarantee the coding performance of key frames for dependent views,a stricter mode decision scheme for P frames is introduced.Experiments show that the method can effectively reduce the computational complexity of depth video coding and save encoding time.During the intra prediction of depth map,it greatly increases the computational complexity of intra prediction to traverse all the possible segmentation patterns of the DMM.In this paper,a fast mode decision algorithm based on an edge classificationon Hadamard transform domain for depth map is studied and implemented.The algorithm uses the results of edge classification by a simple edge classification method to selectively skip the unnecessary DMM during the mode decision process.In addition,based on the fast edge classification method in the original algorithm,this paper optimizes the candidate set of intra prediction from the smoothing blocks and boundary blocks to simplify the decision process of the intra prediction.Experimental results show that the proposed algorithm can achieve considerable coding time savings meanwhile guarantee the coding quality.