Design and application of new multiple description coding schemes

Multiple description coding is a source coding based solution for packet loss problem in communication networks. In this thesis we first present a two-stage algorithm for two-channel multiple description predictive coding. In this algorithm, a predictive encoder is used in the first stage of each description. The second stage is designed to refine the joint reconstruction from the first stage. Theoretical analysis and simulation results show that this method is very efficient in the high rate multiple description coding of strongly correlated sources. We then present a low complexity M-channel multiple description coding scheme which is designed based on two-rate coding and staggered quantization. For correlated sources, a two-rate predictive coding is used in each description. The application of the proposed scheme in lapped transform based image coding is also investigated, and the optimal transform is obtained. Experimental results using both 1-D memoryless sources and 2-D images demonstrate the superior performance of the proposed scheme. Next, a previously developed prediction compensation based two-channel algorithm is extended to generate more than two descriptions. Application of this algorithm in lapped transform based image coding shows a performance competitive to the state-of-the-art algorithms. Finally, this algorithm is further improved using a three layer design and sequential prediction. This improved algorithm is applied to lapped transform based image coding, where the corresponding optimal lapped transform is formulated and obtained. Image coding results show that this method outperforms other latest schemes.