Optimal Predictive Coding of 2 D Fields
José M. F. Moura, Nikhil Balram · 2005
We discuss coding of 2D data using a recursive framework for noncausal Gauss Markov random fields (GMRF) defined on finite lattices. This framework exploits to advantage the structure of GMRFs providing the means to achieve recursive optimal processing, while preserving the noncausality of the field. The compression scheme uses noncausal prediction coupled to vector quantization (VQ). The noncausal prediction fits first a noncausal GMRF to the data, then whitens the data by an inverse filtering type operation, and finally vector quantizes the prediction error field. In this paper, we explain the details of the noncausal prediction. Lack of space prevents us to discuss the parameter estimation algorithm that is needed to fit a :!D model to the data, see [l].