Research On Error Concealment Key Techniques Based On HEVC

With the rapid development of radio television and video entertainmentindustries, video applications tend to be a high definition, high frame rate and highcompression rate. As the current mainstream compression standard, H.264/AVC hasits disadvantages in the future. In April2010, JCT-VT meeting was firstly held inDresden, Germany, which defined a new generation video coding standard calledHEVC (High Efficiency Video Coding). HEVC aims to double the compressionefficiency, which means that the bitrate will be reduced by50percent under the samereconstruction quality for video. However, when video data are transmitted throughIP networks or wireless mobile networks, random error-code and loss of video dataare always appear unavoidably. Since HEVC still adopts the same inter and intraframe prediction mechanism as in H.264, high compressed video data are verysensitive to channel errors. Sometimes even few errors may have serious impact onthe reconstruction quality of video data. Therefore, the error-control technique basedon HEVC is one of important research contents in the application.The basic principle of error concealment is as follows. Since video contentshave much in common both in the time domain and the spatial domain, the wronglyreceived video data can be reconstructed by using its neighbouring data in the timeand special domain which are normally received. This paper closely analyses thecharacters of HEVC, and further studies error concealment technology in the post-decoding process. Then it proposes error concealment algorithms which take thecharacters of HEVC into account. Here is the major content of the paper:Firstly, this paper proposes an error concealment technique based on differentmotion and texture regions. Traditional algorithms do not fully consider the textureand motion features of video content. This paper divides the regions which errorblocks belong to into three types, that is, background regions, simple motion regions and complex motion regions. The blocks in the background regions keep a strongcorrelation in time domain; the blocks in the simple motions region always havesimple structures and strong motion consistency; the blocks in the complex motionsregion always have rich motion information and more accurate concealment isnecessary. This paper adequately considers the motion properties of coded blocks inthe different regions. For error blocks in three different regions, different concealedmethods are adaptively adopted. Experimental results show, compared with manytraditional algorithms, the proposed method can obviously improve the reconstructedvideo quality with low complexity.In addition, this paper also proposes an error concealment technique based onvariable-size blocks for the whole frame loss. When video data are transmitted insome communication channel with poor condition, the whole frame may be lost. Thetraditional algorithm is mainly based on motion vector extrapolation (MVE) withsmall size block. MVE can obtain good reconstruction quality, but it is not suitablefor high resolution video frames because of its high computational complexity.Based on the analysis of temporal correlations of video sequences and quard-treestructure in HEVC, we propose an error concealment algorithm which is applied towhole frame loss in HEVC. The proposed algorithm firstly judges whether the lostblock is a simple motion block according to the correlation between the lost blockand the motion vector of received correctly blocks in the reference frame, then thedepth of lost block is estimated by the depth of received correctly blocks in referenceareas and the correlation of extrapolated motion vectors. At last, the lost block withdifferent block sizes is concealed with MVE algorithm adaptively. Experimentalresults show, compared with the traditional algorithm, the proposed method greatlyreduces computational complexity while guaranteeing high video quality.