Research On Fast Coding Method For Depth Maps Of 3D-HEVC

From live video,video monitoring and video conference to online courses,video technology plays an increasingly important role in people's lives.In order to meet the needs of people to view the same thing from different perspectives,JCT-3V has issued 3D-HEVC standard.As an extension of HEVC,the input video of 3D-HEVC has become videos of multiple viewpoints and their corresponding depth videos on the basis of HEVC,and it has also added a variety of new coding tools for efficient coding,however,the coding complexity has greatly increased,which is not conducive to its practical application.In order to reduce the complexity of depth map coding,this thesis proposes the following two fast depth map coding algorithms,which optimizes the depth map coding algorithm in 3D-HEVC technology.Firstly,a fast decision algorithm of depth map intra mode is proposed.Based on the characteristics of steep edges in the depth map,the edge detection method is used to judge whether the current prediction unit(PU)has a steep edge,and the traversal of depth modeling modes(DMMs)can be skipped selectively;for PU that cannot skip DMMs,the correlation of rough mode selection results and the most probable mode with wedge mode are used,only traverse wedge mode related to angle mode in candidate mode,which significantly reduces the number of wedge modes to be traversed and speeds up the decision process of DMMs.The experimental results demonstrate that compared with standard encoders,the coding time of the depth map is reduced by 35.4% on average.Correspondingly,the loss of average PSNR is 0.29 d B and the increase of average bitrate is 1.06%,respectively.Secondly,a fast algorithm for depth division of Coding Unit(CU)in depth map is proposed.This method selects multiple reference CU and assigns different weight values to them.The sum of the depth value and weight value of each reference CU is calculated to predict the depth range of current CU,and the unnecessary CU depth is skipped.Then,the correlation between texture complexity of CU and CU partition is used,it can be determined whether the current CU is smooth by using the sum of the difference between the two diagonal pixel values and the variance of CU,and terminate the CU partition process in advance.The Experimental results demonstrate that compared with standard encoders,the encoding time of the depth map is reduced by 47.92% on average,Correspondingly,the loss of average PSNR is 0.23 d B and the increase of average bitrate is 0.38%,respectively.