Vector quantization of residual images using self-organizing map

Vector quantization (VQ) is a signal compression technique which can provide high compression rates, and the self-organizing map (SOM) can be employed in the generation of VQ codebooks. Exploiting the ordering property of SOM, the encoding process can be considerably accelerated by using a two-level search. In this paper, we deal with the VQ of prediction error (residual) images in image sequence coding. The results show that the codebooks generated by SOM and the widely-used LBG algorithm achieve almost the same performance, but the encoding process can be realized in a more efficient way by exploiting the ordering property of SOM.