Abstract
We propose an algorithm that improves the performance of rate-distortion-based multiple description coding (RD-MDC). The gain is particularly significant in the high redundancy region, where RD-MDC suffers a major performance penalty with respect to MDC bounds. The improvement is obtained with negligible additional computational cost, by exploiting the coarse information also at the central decoder. The proposed method can be generalized to all MDC schemes that use scalar quantization, without modifying the quantizer structure. This feature guarantees the generation of descriptions that can be decoded without any modification of the decoder.