Error-resilient and progressive transmission and decoding of compressed images

This paper describes an approach for structuring and transmitting a sequence of variable length codewords in order to allow for a robust and progressive transmission over noisy channels. The dependencies within the resulting bitstream can be easily modeled with a stochastic automaton leading naturally to efficient soft decoding techniques with a complexity comparable to the complexity obtained with classical bitstream structures. Simulation results reveal, with both theoretical sources and real images, high performances in terms of symbol error rates and mean square reconstruction error at no cost in compression efficiency with both hard and soft decoding.