Novel vector quantization based algorithms for low-power imagecoding and decoding

摘要

In this paper, a novel scheme for low-power image coding andndecoding based on vector quantization is presented. The proposed schemenuses small codebooks, and block transformations are applied to thencodewords during coding. Using small codebooks, the proposed scheme hasnreduced memory requirements in comparison to classical vectornquantization. The transformations applied to the codewords extendncomputationally the small codebooks compensating for the qualityndegradation introduced by the small codebook size. Thus the coding tasknbecomes computation-based rather than memory-based, leading tonsignificant power savings since memory-related power consumption formsnthe major part of the total power consumption of a system. Since thenparameters of the transformations depend on the image block underncoding, the small codebooks are dynamically adapted to the specificnblock under coding leading to acceptable image qualities. The proposednscheme leads to power savings of a factor of 10 in coding and of anfactor of 3 in decoding, at least in comparison to classical full-searchnvector quantization. The main factor affecting both image quality andnpower consumption is the size of the codebook that is used