Research On Error Concealment Techniques Of Stereoscopic Video

With the acquirement of the video quality ,stereoscopic video with high definition and multi-view will be the trend. Compared with the 2-D video , double data of muilti-view video will be the problem in store and transmission over internet and how to efficiently compress multi-view videos is the hot research topic of multi-view video coding technique . In addition, robust video delivery via error-prone networks is an important application, which can be effectively dealt with using error concealment techniques. Error concealment has been widely exploited at the decoder side of traditional single view coding systems. However, the error control of multi-view video coding is still the research spot. this paper carries on the research of error concealment process on stereoscopic video coding to develop a more robust multi-view video delivery, and proposes a novel error concealment technique for stereoscopic video coding.Firstly, the thesis summarizes the problems of the error concealment of stereoscopic video based on the coding structure of multi-view video. Analysis is conducted on the statistical of the macroblock mode relativity between the current frame and the reference frame and the regular pattern of macroblock mode relativity.The conclusion is that the macroblock mode relativity between the current frame and the B frame reference piciture is higher and the trend of macroblock mode relativity in view0 is the same as that in view1. Different block size is appled in different texture.Secondly, a novel fast algorithm for reference frame selection that select a more reliable reference frame between foreward and backward reference frame is proposed based on the statistics of the macroblock mode relativity and the trend of the macroblock mode relativity.We proposed a novel error concealment algorithm of entire right frame losss in tereoscopic video based on the macroblock mode selection and reference selection. Simulations show that PSNR can be improved by0.42-1.31dB compared to the frame copy algorithm.Thirdly, this paper proposes an error control technique based on global disparity vector for stereoscopic video coding and proposes an inter-view error concealment mode based on the global disparity vector model and the macroblock mode and reference selcection for inter-view error concealment technique.Simulation results show that the proposed technique indeed disentomb the inter-view correlation and PSNR can be improved by0.12-1.87dB compared to the frame copy algorithm. Finally,the thesis summarizes the work done and describes possible directions for future studies.