A nearly lossless vector quantization algorithm for compression of remotely sensed images

Abstract

Most compression algorithms are designed for minimizing a squared error criterion. The squared error criterion does not accurately represent the fidelity requirements for scientific image compression. In this paper we propose a distortion measure which correlates with subjective evaluations, and an adaptive vector quantization algorithm which minimizes this distortion measure. A new approach to codebook design is presented to replace the nearest neighbor approach.