Modeling and recursive state estimation for two-dimensional noncausal filters with applications in image restoration

Noncausal blurring filters are frequently encountered in applications of 2-D signal processing. In this paper we obtain low-order, equivalent, quarter-plane (QP) causal Roesser separable in denominator state-space models for such noncausal blurring filters. We utilize the equivalent QP causal models to develop a minimum-variance deconvolution (MVD) algorithm, which can be used in image restoration and other 2-D MVD problems.