Recovery of block-coded images from channel errors

A method to recover good-quality images from severe transmission errors in a block coding environment is proposed. In block-coded images, channel errors usually result in annoying blocky noise. The proposed scheme tries to remove or conceal this blockiness through error detection and block replacement. The single bit-error dominance in binary transmission and high pixel correlation of common images are utilized. Specifically, assuming that the boundaries of image blocks in the received bit stream are detectable, all the bit patterns at unit Hamming distance from the received one are first decoded into candidate image blocks. Then, the candidate image block with minimum gray-level transition across the block boundaries is selected for final recovery. This method works not only on fixed-length binary codes but also on variable-length binary codes as long as block boundaries are known to the receiver. The comparative performance and possible improvements of the scheme are studied