Side-Match Vector Quantization for Reconstruction of Lost Blocks

Abstract An ordinary VQ has no ability to reconstruct the blocks of lost indices which are caused by the lost packets or the damaged storage. This paper describes a new side-match vector quantization (SMVQ) algorithm to recover the blocks of lost indices in an ordinary vector quantization algorithm. It can predict the lost block from the neighboring blocks of this lost block without any side information. The SMVQ algorithm tries to make the gray level transition across the boundaries of the vectors as soon as possible. In our experiments, the improvement using SMVQ to recover the lost blocks is up to 3.618 dB for the image Lena. An interleaved SMVQ (ISMVQ) algorithm is also proposed in this paper. The ISMVQ algorithm combines the VQ algorithm and the SMVQ algorithm. In our experiments, the improvement of ISMVQ over VQ is up to 1.488 dB at the same bit-rate for the image Lena.