A family of fast index and redundancy assignments for error resilient multiple description coding

Abstract

Multiple description coding (MDC) was originally developed to combat channel failures and packet losses. For the particular case of fixed-rate scalar quantization, it has been shown that the multiple description scalar quantizer (MDSQ) can be made robust to binary transmission errors, through a proper binary labelling of the indices transmitted in each description. The design of this so-called index assignment (IA) is computationally very demanding, especially for systems with large redundancy or high bit rates. In this paper, we propose an alternative low-complexity IA design for this problem. For low redundancy, this new design provides results that are close to optimum (in the sense of minimizing the side distortions for a given central distortion), while maintaining the robustness to transmission errors and being amenable to designing systems with large redundancy or high bit rates. For high redundancy, the proposed design provides high robustness while allowing for successive decoding of different signal qualities. Thus, it can be applied for the case of progressive quantization and transmission.