Research On Fully Scalable Wavelet-based Video Coding And Post-Processing Techniques

Being one of the most important information exchange technology in the era of network,visual communication will be widely applied in different markets. The heterogeneity ofnetwork infrastructure and terminal device brings some big challenges to the video serverto provide the service of multimedia transmission. To be an effective method withperformance of memory saving and reduction to computational complexity of server, thescalable video coding has been extensively studied and applied. On the other hand, theproblem of improving the image quality at low bit rate deserves to be studied because thecoded images or video take on some serious coding artifacts under the constraint of thebandwidth of the network or satellites. For the purpose of realizing the fully scalable videocoding and improving the qualities of coded images at low bit rate, the dissertationconducts researches in the following two aspects: 1) In the framework of fully scalablevideo coding with motion compensated 3D wavelet transform, the corresponding codingtechnologies including adaptive time filtering, independent subband coding and rate controlare studied. To reduce the computational complexity of the adaptive arithmetic coding, theprobability model representation problem is studied. 2) To remove the artifacts of ring andblurring produced from the wavelet codec at low bit rate, the post-processing methods arestudied to improve image qualities. The dissertation studies the fully scalable 3D wavelet video coding with motioncompensation algorithm. The code stream structure satisfying three scalable properties inthe same time is provided . Then, the adaptive time filtering is studied including thefollowing three aspects: 1) Classifying pixels according to the connection property frommotion estimation, an adaptive filtering method which disposes the connected andunconnected pixels is presented. 2) An adaptive time filtering formula is provided to avoidthe presentation of virtual image by explaining the time scalable decoding process. 3) Themotion vector drift is interpreted under the resolution scalable decoding. Also, thetime-space frames are encoded with the independent 3D subband embedding bit-planecoding. Among the coding method, a quad-tree decomposition skill is used to express thevariable size zero block in order to improve the performance of the significant map coding.The high order context arithmetic model is applied to approximate the entropy limit in theprocess of arithmetic coding of the output information of the bitplane coding. As to achievethe optimal performance of the fully scalable 3D wavelet video coding method, thedissertation studies the rate control technique of the coding method. The rate control isdivided into two parts: rate control among the multi-resolution time-space frames and bitallocation among the subbands in a frame. According to the propagation of distortion in theprocess of time filtering, the distortions of the original time frames are converted to thoseof the time-space frames. Baed on this analysis, the rate control algorithm among thetime-space frames is presented for the purpose of controlling the reconstructed videosmoothness between the neighboring frames. Based on the properties of the theoretical IIIrate-distortion curve, the optimal bit allocation among subbands algorithm is presentedwith the practical normal rate-distortion curve by deleting some rate-distortion points.Experiments proves the rate control algorithm not only maintains the smoothness of thereconstructed video, but also improves the objective PSNR value effectively. Experimentsalso show that the fully scalable wavelet video coding has both the full scalability and goodcoding efficiency, so it can be applied to the streaming media service. The dissertation studies the probability model representation of the adaptive arithmeticcoding. According to the detachable character between the probability interval map...