New approaches for space-invariant image restoration

The problem of restoring images degraded by space-invariant blurs and noise is addressed. Two approaches, one based on Kalman filtering and the other on projection onto convex sets (POCS), are proposed. The Kalman filtering approach modifies the image model used in the usual reduced-order model Kalman filtering (ROMKF) approach to obtain a more accurate representation of the image distribution. The proposed POCS-based approach utilizes novel space-domain constraints defined in terms of the space-varying blur function. Both approaches have been shown to effectively restore images degraded by LSV (linear space-variant) blur functions in the presence of additive noise.<>