Balanced multiple description scalar quantization

Abstract

This paper tackles the problem of the generation of an arbitrary number of balanced descriptions with multiple description scalar quantization (MDSQ). We show how, with a very low complexity, we can vary the number of descriptions and the redundancy between them, in order to adapt to different channel characteristics. A comparison with state-of-the-art MDSQ schemes shows a better performance of our solution in terms of an average distortion at the receiver, which comes from the flexibility of our solution to better adapt to various lossy conditions.