Coding characteristics of the pixel-adaptive DPCM utilizing the variable-length-codes

Abstract

This paper discribes the coding characteristics of the modified DPCM for data compression of still images for digital still cameras. In pixel adaptive DPCM, the pixels are classified into eight groups according to the magnitude of the prediction error of preceeding pixels, and the prediction error signals of each group are quantized using eight nonlinear quantizers, each of which is designed optimally to each group. The performance of the pixel-adaptive DPCM can be improved by utilizing the variable bit-rate scheme. It is shown that for “girl Y” the S/N ratio is improved to 48db by 6db at 3.0bits/pixel, that this improvement is due to the relief of the slope overloading at the edges, and that the spectrum of the coding errors is distributed uniformly in the 2-D spatial frequency region. © 1991 IEEE