Multiple description coding using transforms and data fusion

This paper proposes a new multiple description coding technique that uses different data transforms to generate multiple descriptions of a given signal. A linear fusion is employed to reconstruct the signal as the output of a central decoder. This new method exhibits the advantages of multiple description vector quantization while having the complexity of multiple description scalar quantization. Another technique called "forced errors in channels" is then introduced to reduce the distortion in the central decoder output allowing some increase in the side distortions. When applied to images, the new techniques yield results, which are competitive compared with the results from current state-of-the-art coders.