Boundary matching detection for recovering erroneously received VQ indexes over noisy channels

This paper addresses the issue of transmitting and reconstructing vector quantization (VQ) coded images over a noisy channel. It presents a novel approach which exploits the spatial contiguity and interpixel correlation of image data sequences. Specifically, a boundary-matching-based detection algorithm (BMDA) is proposed along with a principal-component-splitting-based indexing (PCSBI) scheme which organizes the codebook in a way similar to the GLA-splitting-rule-based indexing scheme, so that some bits of an index are more important than others. Simulation results show that the proposed scheme yields significantly better visual quality and higher signal-to-noise ratio (SNR) than a maximum a posteriori (MAP) detection-based scheme, and has robust performance. It is also shown that incorporation of PCSBI with BMDA greatly reduces the detection complexity, thus facilitating the implementation of BMDA.