Many current applications such as data hiding and watermarking can be posed as the problem of channel coding with side information. There has been considerable interest in designing codes to try and attain the theoretical capacity of the problem. In order to achieve capacity, a powerful channel codebook that partitions into a powerful source codebook should be chosen. The data to be embedded will index the source codebook partition. The constructions that exist in the literature, however, are typically based on powerful channel codebooks and weak source codebooks and hence remain at a considerable gap to capacity. In this project, we present two methods of construction that are based on a powerful channel codebook (i.e., turbo codes) and powerful source codebook partitions (i.e., trellis coded quantization) to try and bridge the gap to capacity.
1Visiting Professor, University of Michigan