Two‐dimensional interpolative DPCM
Shiro Handa, Hatsukazu Tanaka · Electronics and Communications in Japan (Part I Communications) · 1988
Abstract Differential PCM (DPCM) has often been used for encoding of images, but its efficiency is very poor at low transmission rates. Discussion of DPCM is based on an autoregressive (AR) model which is causal in the case of raster scan sampling. However, originally, images do not satisfy causality in this sense, and a better description of the two‐dimensional image information can be obtained by an interpolative model representing the relations between neighboring pixels. This paper proposes a two‐dimensional interpolative DPCM which is based on such a model. This method has the following special feature: if the output samples of a DPCM process based on a causal AR model are regarded as intermediate decoded values, then the final decoded values are yielded by a filtering process based on an interpolative model. The results of simulation using actual images demonstrate SN ratio improvement at a low bit rate (1 bit/pel) and also the stability of encoding characteristics in the case of quantitative step‐size variations. It is shown that fixing of the prediction and interpolation coefficients results only in slight deterioration.