Study On Stereoscopic Video Error Concealment Technology For Entire Frame Loss

Sterescopic video coding and transmission is an important study in the field of video signal processing.Currently,the most wired and wireless networks can't provide QoS guarantee,so the error and toss of video data are inevitable during sterescopic video stream tansmission.So,the study of sterescopic video error concealment algorithm has become more important in the sterescopic video processing techniques investigation.The error concealment techniques for entire right frame loss in sterescopic video are researched in this dissertation.The research is based on the newest multiview video coding standard(JVT MVC) which is an extension of H.264/AVC.First,an error detection measure is implemented in the JVT MVC Reference software JMVM6 by the analysis of bit stream structure in the JMVM,which lays a good foundation for the study of error concealment algorithms.Second,an existing error concealment algorithm dealing with the loss of entire right frame is implemented in JMVM6,the existing algorithm is based on temporal relativity.According to the analysis of the existing algorithm,a refined error concealment algorithm based on the temporal relativity is implemented in this paper.The experiment results show that the refined algorithm can achieve 0.2～2.79dB average PSNR gains for recovered images compared with the original algorithm for the gentle sterescopic video sequences.Finaly,this paper analyzes the relativity between the two views according to the existing coding scheme of stereoscopic video sequence,then proposes an error concealment algorithm based on the inter-view relativity for the entire right frame loss:MSkip which is derived by improving the existing error concealment algorithm for Coarse Granular SNR(Quality) Scalability(CGS) enhancement layer in monocular video.Simulation results show that,for the sterescopic video sequences which have small global disparity,the MSkip algorithm can achieve 0.35～3.23dB average PSNR gains compared with the existing algorithm based on temporal relativity and 0.24～6.45dB average PSNR gains compared with TD algorithm in JMVM.