Research On Predictive Coding In 3D-HEVC

Three dimensional video(3DV) is one of the research points in the near future in the field of video coding that can be immersive stereo feeling for the audience. In consideration of advantages with backward compatibility and supporting interaction of the multi-view video plus depth(MVD), the motion picture expert group chooses MVD for the new generation of 3DV standard data format. Due to that the multi-view video and multi-view depth information must be coded and transmitted to decoder end, it makes the rate increase significantly. Therefore, it is particularly important to improve the coding efficiency of 3DV. 3D-HEVC is being hatched MEPG standard that researched on MVD on the basis of HEVC. 3D-HEVC put forward too many coding tools on account of the characteristics of 3DV that add about 50% of the rate to provide three-dimensional visual experience. This paper focuses on fast coding technology based on correlation of inter-view for 3DV and advanced residual prediction technology which can improve the coding efficiency for 3DV. The main contents are followed:Advanced prediction residuals technology is an efficient technology for 3D video coding which aims at exploiting the residual correlation between views. It can effectively reduce the coding rate of non-base view. Its key point is structure the predictive value of residual for the current residual through using the existing motion information. If the corresponding blocks of base-view using the intra blocks, then it does not contain motion information, which limits the use of advanced residual prediction algorithm in a certain extent. Focusing on the problems above, an optimization algorithm is put forward in this paper. The algorithm uses neighbor block’s motion information of intra blocks instead of intra blocks’ zero vector. It can improve the prediction accuracy and coding efficiency of ARP. Experimental results demonstrate that the proposed algorithm can reduce the coding bit-rate by an average of 0.12% of non-base view.The video at the same time between the different views has high correlation in the texture for the multi-view video. 3D-HEVC employs careful partition algorithm of prediction unit to improve coding efficiency. In order to reduce the computational complexity, an inter prediction unit optimization algorithm is put forward in this paperthat based on the correlation of inter-view. According to the correlation of PU modes between inter-view pictures, the candidate list of prediction modes is constructed for current PU and the PU candidate modes reduced. In the transform coding of 3D-HEVC, TU uses the residual quad-tree(RQT) transform structure. Rate-distortion cost of all the different RQT depth are compared transversally to get the optimal transform unit. The coding efficiency is improved significantly by the new transform technology. However, computation complexity is also increased compared to previous standard. But the ARP makes full use of the correlation of residual that can reduce the energy of residual. After experimental statistics, it is found that about 97.38% of TU blocks using advanced residual prediction technology have not divided. Base on this, TU early termination partition optimization algorithm is proposed this paper by means of defining the ARP flag. The experiment results show that the two proposed methods can save coding time of about 32.2% on average for non-base texture compared with the 3D-HEVC common test platform under the condition of minimal lose of coding efficiency and PSNR.