Research On Key Technologies Of Depth Video Coding In 3D-HEVC

3D video technology has been developing rapidly in recent years and lots of researches have been carried out on 3D video.In 3D video coding,the coding method for multi-view video plus depth(MVD)format is the current hot topic.MVD videos include multi-view texture videos and corresponding depth videos.At the receiver side,the MVD based video system can generate arbitrary intermediate views by means of depth image based rendering(DIBR).Thus,depth video coding is important in 3D video coding.High Efficiency Video Coding(HEVC)achieves better coding efficiency compared with the conventional H.264/AVC.3D-HEVC,as the 3D extended version of HEVC,has proposed corresponding coding tools based on the characteristics of MVD videos.However,these tools lead to unaffordable computation complexity for intra prediction modes and the overall coding efficiency still needs to be further improved.For this purpose,this paper carries out research on depth video coding in 3D-HEVC.For better prediction of edges in depth maps,new depth intra modes were proposed,such as Depth Modeling Modes(DMMs).DMM Mode 1(Wedgelet mode)is a method modeling the depth block with straight line,whose kernel is to find the best partition pattern for a depth block.Conventional algorithms are based on blind traversal and lead to unaffordable computation complexity.To address this problem,a simplified Wedgelet search algorithm based on the similarity between the edge in an actual depth block and the separation line in the best matching pattern,is proposed to reduce the searching complexity.Experimental results show that considerable encoding time saving is achieved with negligible performance loss on synthesized views.For the depth edge blocks,conventional intra and inter prediction modes still need to be improved.Therefore,a combined intra-inter prediction for depth video coding is proposed in this paper.Firstly,the best intra and inter modes of the current depth block are obtained.Then,the two modes are assigned to each partition region of the current edge block.Finally,the weighting coefficients of each region are adjusted adaptively to realize the combined prediction.Experimental results show that the proposed method achieves a better coding performance compared with the conventional prediction modes.