Variable‐length‐code‐selective reversible predictive coding for multilevel images

Abstract

AbstractTo construct efficiently an image database system for high‐tone images, a variable‐length, code‐selective reversible predictive coding scheme with a high‐compression ratio is proposed. This coding scheme is based on application of the quantizer‐selective method in the DPCM coding [11] (previously proposed by the present authors) to an adaptive selection of a variable‐length code. This is an adaptive coding scheme in which a suitable code is selected from the pre‐designed variable‐length codes for each pel. This code selection method requires no additional information, since the code selection is carried out based only on information of pels for which the coding has already been completed. The predesigned variable‐length codes and the threshold values for the code selection are determined by a model of the probability distribution of transformed prediction‐error signals. The initial setting of coding parameters is constant and independent of the image to be encoded. This scheme has, therefore, much generality that is independent of any kinds of images. The coding rate of this scheme is improved by about 0.3 to 0.5 bit/pel compared with the optimal nonadaptive coding scheme using a single Huffman code. It was confirmed that the proposed scheme has the highest coding efficiency compared with several conventional representative reversible coding schemes.