Enhanced trellis coded quantization scheme for robust image transmission

In this paper, we present an enhanced scheme over a trellis coded quantization (TCQ) approach recently proposed for the transmission of compressed images over noisy channels. This original scheme employs block DCT, TCQ, and fixed length bit allocation so that the compressed data is relatively insensitive to channel noise. Although the fixed length coding is able to avoid the catastrophic error propagation, one problem of the approach is the severe distortion of the whole blocks in the case of high channel error rate. To reduce the unpleasant error block effects, we propose to re-arrange the subsources of the image before the process of TCQ. With such a preprocessing, the degradation caused by the channel error can be distributed to the whole image, instead of within a particular block. Preliminary results show that this enhanced TCQ scheme is able to improve both the PSNR and the visual quality of the reconstructed image.