Research On Multiple Description Image And Video Coding

Network congestion and delay sensibility pose great challenges for multimedia communication system design. This creates a need for coding approaches combining high compression efficiency and robustness. Multiple description (MD) coding has emerged as an attractive framework for robust transmission over unreliable channels.This thesis aims to design the schemes of effective MD image and video coding. For image coding, the two schemes, that is, MD image coding based on lattice vector quantization and MD image progressive transmission based on shifted lattice vector quantization are proposed respectively. For video coding, the two schemes, that is, MD video coding based on rate-distortion optimized temporal sub-sampling and MD video coding using priority transmission are presented respectively. The details are as follows.â‘ The classical theory of MD lattice vector quantization (MDLVQ) is employed to design image codec. In view of the characteristics of wavelet coefficients in different frequency subbands, MDLVQ is applied in an optimized way, including an appropriate construction of wavelet coefficient vectors, the optimization of MDLVQ encoding parameters such as the choice of sublattice index values and the quantization accuracy for different subbands. More importantly, optimized side decoding is employed to predict lost information based on inter-vector correlation and an alternative transmission way for further reducing side distortion. Compared with MD image coder based on scalar quantization, at the same bit rate, the central distortion and side distortion of the proposed scheme have 0.2-0.7 dB and 2-5 dB improvement respectively. Compared with MD image coder based on correlated transformation, our scheme can also improve around 2 dB in central distortion and 5 dB in side distortion.â‘¡According to the geometrical structure and the special relationship of lattice vector quantizers, a MD quantizer called shifted lattice vector quantization (SLVQ) is employed in MD image coding to realize progressive transmission over unreliable channels. Compared with SVS scheme, the gap between central and side distortion has been reduced from 9-11 dB (N = 7) to 4 dB. In view of the characteristics of wavelet coefficients in different frequency subbands, the algorithm of modified zerotree coding is also applied to improve compression performance. Compared with the other schemes based on MD scalar quantization for progressive transmission, at the same bit rate, the central distortion in the proposed scheme has achieved 1-1.5 dB improvement and the side distortion has 0.2-0.5 dB improvement.â‘¢An effective MD video codec is designed based on rate-distortion optimized temporal subsampling. In the scheme, Lagrangian optimization method is applied in the mode choice for frame interpolation and duplication. And the scheme has perfect compatibility with the standard source codec, such as H.26X and MPEG-X. At the same bit rate, the proposed scheme has taken 0.4-0.8 dB advantagement in side distortion over the conventional temporal scheme. Especially, for the frames containing high motion, almost 8 dB improvements in side distortion can be achieved by the proposed scheme. In packet loss network, the proposed scheme has also obtained 0.3-1 dB compared with the conventional one. Compared with the 1D spatial subsampling based MD coder, the proposed scheme has 0.2-1.4 dB improvement in central distortion and 0.3-1.2 dB in side distortion. Compared with the 2D spatial subsampling based MD coder, the proposed scheme has 2.5-3 dB improvement in central distortion and 2-3 dB in side distortion.â‘£An effective scheme of multiple description video coding is proposed for transmission over packet loss network. In this scheme, we attempt to overcome the limitation of specific scalable video codec and apply FEC-based multiple description to the standard video codec. According to different motion characteristics between frames, an original video sequence is divided into several sub-sequences as messages, so in each message better temporal correlation can be maintained for better estimation when information losses occur. Flexible GOP (Group of Picture) organization is applied to match the motion between frames. As a result, the frames with high motion changes can be encoded as I frames and reasonable priorities can be assigned in each message. Furthermore, the priority is designed in view of packet loss rate of channels and the significance of different segments in bit streams. At high packet loss rate, the proposed scheme can degrade gracefully while the equal protection has a sharp transition. Here, the largest gap that the propose scheme surpasses the equal protection is about 8 dB.