Research On Error Concealment For Three Dimensional Video Transmission

The rapid development of multimedia technology greatly enriches people’s lives, facing thediverse world, people pursue more high-quality visual experience. Three-dimensional video hasattracted wide attention from academia and industry with its outstanding features such as flexibleinteractivity. Efficient compression encoding technology has reduced the error resilienceperformance of bit stream, and the bit stream is very sensitive to channel errors. It is prone toproduce packet loss when the compressed bit stream transmits over a network channel. As a result,the bit stream can’t be decoded correctly in the decoder, seriously affecting the quality ofexperience(QoE). So, the error control technology is more and more important. In view of theabove problems, this dissertation focuses on the error concealment technology in three-dimensionalvideo transmission. We use the temporal and spatial correlation existing in the stereo videosequence, combining the content features of video images, to improve the viewing quality of thevideodecoderas faras possible. Themainresearchcontents ofthispaperareas follows:(1) This paper proposes an error concealment algorithm of right view part macro block(MB)loss for stereoscopic video IPPP coding prediction structure. Motion and motionless maps in thelost frame of the right view in stereoscopic video are estimated, According to temporal correlationbetween adjacent frames of the lost frame and the spatial correlation between viewpoints. The lostMBs are divided into motion MBs and motionless MBs. According to different loss block, weadopt different error concealment methods, to avoid using the same method for all the lost blocks.Experimental results show that the proposed algorithm can efficiently improve the subjective andobjective quality of the reconstructed frame. PSNR of concealed frames are improved about0.77–3.22dB on average compared with concealment method in JM under different quantizationparametersanddifferent macroblock lossrates.(2) This paper presents an error concealment method of entire frame loss. The lost frame isdivided into the background and foreground motion area according to the information betweenbidirectional reference frames of the lost frame. The static background regions can be simplyrestored by copying the corresponding information in temporally reference frames; For the lostblocks in foreground motion region, we combine the temporal motion compensation and spatialcompensation to restore the loss information. In the experiment, we first test the effectiveness ofregion segmentation for the time-domain bidirectional motion compensation and disparitycompensation, the results show that the PSNR of three test sequences can be improved by about0.2dB~8dB. Then we test the results of proposed algorithm, the results show that the proposedalgorithmcomparedtothecomparison algorithmcan improves0.32dBto5.09dB. (3) Because of drawbacks of the traditional restored image quality assessment, we point outthat restored image should improve QoE. For many classical error concealment methods and theerror concealment algorithm proposed in this paper, we assess restored images using differentobjective assessment methods. To further reflect the quality of the user experience, we conductsubjective experiments on right view restored images and stereoscopic images. Experimentalresults showthat theproposed errorconcealment algorithmismorein linewithuserQoE.