Research On Low Complexity Algorithm In 3D-High Efficiency Video Coding

With the accelerative development of multimedia industry,each filed has put forward higher requirements for definition and resolution of video.Three-dimensional(3D)video has brought new feeling experience to audiences with its unique sense of stereoscopic and interactivity,which has been applied to many fields.However,the amount of data generated by 3D video is a great burden on network transmission and storage resources,limiting its application in real-time applications.The coding and compression of 3D video has become the focus of attention at home and abroad.Therefore,the video coding joint coding group has published the 3D-high efficiency video coding(3D-HEVC)standard to compress current 3D video data.3D-HEVC is the 3D extension of HEVC/H.265 coding standard,continues to adopt the important coding techniques in HEVC,including coding unit,prediction unit and motion vector tools.In addition,it also uses disparity vector,inter-view prediction and view synthesis prediction as coding techniques.The new tools bring about the enhancement of coding efficiency and compression rate,and result in the increase of coding complexity as well.In order to futher reduce the time complexity of video coding and enhance 3D video in the real-time applications,this paper mainly finds out the encoding rule and theoretical basis from the prediction module,and designs a low complexity algorithm of 3D-HEVC based on the original coding platform.Firstly,for the texture map coding,we analyse the coding modes in texture map and summarize the different characteristics between the inter-frame prediction of Merge and other Inter modes.Accoding to the motion vector and disparity vector of prediction unit candidate list building process,the correlation candidate value of spatial,temporal and inter-view can remove redundant candidates adaptively,which reduces calculation of inter prediction and improves the encoding efficiency.In addition,if other Inter modes meet the customized conditions,the module of motion estimation and disparity estimation will be skipped as well,which results in the encoding process is further compressed.The experimental results show that the algorithm can save 33.03%of the coding time of the texture map.Secondly,for the depth map coding,we study the coding similarity between texture map and depth map and use the similarity to predict the coding of depth map,including the partition of coding unit,prediction mode and so on.When the texture coding is completed,the encoded information is saved to the buffer memory,the corresponding depth map can invoke the encoded texture information directly,and skip the remaining partitioning of coding unit,which achieves the fast decision of coding unit and prediction unit.Based on the relationship between texture map and depth map,the fast coding algorithm of depth map can save time complexity of depth map by 41.89%when the coding distortion decreases minimal.