Multiple description scalar quantization with successive refinement

This work presents a framework for multiple description coding (MDC) based scalable image coding. The paper starts with defining the conditions for deviating from the conventional multiple description scalar quantization (MDSQ) by extending the MDSQ into multiple-channel (more than 2) unbalanced descriptions. These two conditions are extended to define the new framework for MDSQ with successive refinement. We propose that the split or refinement factor for each of the refinement side quantizers should be greater than the number of diagonals filled in the index assignment matrix of the side quantizer design and should not be integer multiples of each other, in order to reduce the distortion in joint decoding when the number of descriptions included in joint decoding is increased. Proofs and verification by simulations of these propositions are are shown in the paper. Experimental results show high robustness of the proposed scheme compared to single description scalable image coding.