Error-resilient transmission of compressed images over very noisy channels using soft-input source decoding

In this contribution we utilize an optimal estimation approach for error-resilient transmission of compressed images over AWGN channels. In contrast to other methods, we mainly rely on the implicit residual redundancy inherent in the subband images and the bit-reliability information at the channel output for error protection. It is shown that by including a priori knowledge about the spatial correlations in the subband images, we obtain a highly increased quality of the reconstructed image. This holds especially for very noisy channels and a mean-square estimation of the subband source coefficients.