Edge modified vector quantization for image compression

Abstract

A new adaptive source coding technique called edge modified vector quantization (EMVQ) is introduced. By using EMVQ, an image is encoded in two stages: coarse image coding and edge modification. The redundancies in the smooth area are further reduced by applying large block segmentation, hierarchical VQ and finite state prediction. Sharp edges are well preserved by compensation of their encoding quantization error. A high compression ratio was achieved and the reconstructed images still had satisfactory perceptual quality. The EMVQ coding results are demonstrated.<>