Research On Error Resilience Video Encoding And Error Concealment

Due to the broad application of video encoding and communication technology in digital TV, surveillance, video conferencing, and telemedicine, it is one research hotspot concerned by research institutions and scholars. Packet switching network can not ensure enough bandwidth and limited delay at the same time, it’s up to the encoding and transmission system to handle the video quality reduction caused by packet loss and delay in the unreliable network. Furthermore, the development of new technologies like 3-dimensional video and HEVC has brought new challenges.The research on error resilience video coding and error concealment algorithm in this thesis is performed under such background. The details can be described as follows:(1) In order to overcome the problem that the simple mechanisms of redundant slice generation in current redundant coding algorithms can not adapt to adverse network conditions and varied video contents, a redundant coding scheme based on adaptive redundancy allocation, mode and parameter selection, and quantization parameter adjustment is proposed. A flexible redundant picture number decision strategy is firstly used to adaptively allocate redundancy and optimize the target end to end rate distortion cost. Then, quantization parameters of different frames are adjusted based on the distortion propagation and the distortion property of different frames. Lastly, the structure and mode of redundant coding is improved, and a statistical model is proposed to reduce its computational cost.(2) In order to overcome the problem that the robust HEVC video communication algorithm on the error prone channel is lacked, and current multi-hypothesis motion compensation algorithms can not well adapt to varied network conditions, a coding unit level HEVC error resilience coding scheme based on the multi-hypothesis motion compensation is proposed. The error resilient mechanisms of HEVC is fully explored, and a coding unit level multi-hypothesis coding approach with flexible reference frame, number of hypotheses, coding unit mode, weighting factors and coding tree unit structure is proposed, and The distortion attenuation is also analyzed for the proposed strcture to optimize the mode selection with the total influenced distortion.(3) In order to overcome the problem that the video imformation loss caused by packet loss and long delay reduce the quality of the synthesized view in 3-dimension video communication, a multiple description coding scheme of 3-dimension video based on the end to end rate-distortion optimization of the synthesized view is proposed. Firstly, the end to end distortion estimation strategy of the virtual view is derived for the multi-view video plus depth format 3-D video, by analyzing the influence of the loss in the texture and the depth map on the distortion of the virtul view. Then, the optimized redundant coding method is performed to generate descriptions with the estimated end-to-end virtual view distortion. Lastly, by taking the feature of texture and depth map into consideration, the quantization parameter of the depth map is adjusted at the marcroblock level to improve the coding efficiency.(4) In order to overcome the problem that the error resilient mechanisms of current error resilient coding algorithms are single, and have defects in end-to-end distortion estimation and redundancy assignment, a content based error-resilient video coding scheme is proposed, adaptively selecting intra refresh, reference frame selection, or redundant picture coding. Firstly, a reliability tracking strategy is utilized to analyze the end-to-end distortion when complex encoding parameters are used. Then, by analyzing the statistical feature of different error resilience mechanisms, the implementation of these mechanisms is respectively proposed and adaptively selected. Lastly, a content based encoding parameter selection strategy based on motion and texture information is used for redundancy assignment.(5) In order to overcome the problem that the current error concealment algorithms mainly focus on the single description coding bitstream, an error concealment scheme for the spatial multiple description coding bitstream is proposed. The difference between neighbor frames and the residual energy of complementary descriptions are used to estimate the pixel and motion difference distortion of the current macroblock. Then the pixel and motion difference distortion are used to classify macroblocks, and different error concealment methods are chosen for different kinds of macroblock.