IMAGE MODELING USING INVERSE FILTERING CRITERIA WITH APPLICATION TO TEXTURES

Statistical approaches to image modeling have largely relied upon rand om models that characterize the 2-D process in terms of its first- and second-order statistics, and therefore cannot completely capture phas e properties of random fields that are non-Gaussian, This constrains t he parameters of noncausal image models to be symmetric and, therefore , the underlying random held to be spatially reversible, Recent resear ch indicates that this assumption may not be always valid for texture images. In this paper, higher- than second-order statistics are used t o derive and implement two classes of inverse filtering criteria for p arameter estimation of asymmetric noncausal autoregressive moving-aver age (ARMA) image models with known orders, Contrary to existing approa ches, FIR inverse filters are employed and image models with zeros on the unit bicircle can be handled, One of the criteria defines the smal lest set of cumulant lags necessary for identifiability of these model s to date. Consistency of these estimators is established, and their p erformance is evaluated with Monte Carlo simulations as well as textur e classification and synthesis experiments.