In this paper, we present an image decompression scheme that emphasize
s visual perception rather than objective signal integrity between the
orginal and the reconstructed images. The decoding scheme is develope
d in the context of vector quantization (VQ). The technique includes a
projection based image recovery algorithm as part of the decoding pro
cess. The quantized blocks are classified into edge and shade blocks.
The edge blocks are filtered using optimized templates while the shade
blocks are left unaltered. The method can be considered as a constrai
ned smoothing of edges, which does not cause any degradation to the sh
ade (background) regions in the decoded image. The result is a reconst
ruction that is subjectively better than a simple look-up table decodi
ng.