REALIZATION OF 2-D NONCAUSAL MARKOV MODELS FROM CORRELATIONS.

Markov Random field models of 2-D data are useful in various image processing and signal processing algorithms like image coding, image restoration, texture classification and spectrum estimation etc. Since there is no preferred ordering of data in 2-D it is more appropriate to assume that the observation at a point s is dependent on neighboring observations in all directions leading to a noncausal representation. The 2-D noncausal Gaussian Markov Random Field (GMRF) model characterizes the statistical dependency among the neighboring observations by obeying given requirements.