DPCM Coding System for Image Signals Using Markov Property of Prediction Error Signal

Abstract

A real-time DPCM monochrome image coding system, which skillfully utilizes quantized prediction error signal characteristics represented well by the first order Markov process and can transmit image signals more efficiently than an ordinary non-adaptive DPCM system, is investigated.
In this system, two dimensional prediction error signals are divided into two parts. One consists of the large error signals corresponding to the edge part in the picture, the other consists of the small error signals corresponding to the flat part. An information for discriminating these two parts, which are encoded separately using block codes, is transmitted by runlength coding.
Both an analysis based on the Markov model and a simulation using standard images indicate that the average information rate and the average transmission rate of this DPCM system are about 0.2bit/pel lower than those of the non-adaptive DPCM system; besides, these two coding systems can reconstruct perfectly equal images.