NONRECURSIVE FILTER FOR EDGE PRESERVING IMAGE RESTORATION.

The development of a nonrecursive filter for edge preserving image restoration is considered. The original image is represented by a Gaussian Markov random field (GMRF) model. This assumption forces the restoration filter to be a function of GMRF model parameters. Since the original image is rarely available, methods are developed for the estimation of model parameters from the degraded image. The degradation is due to signal-independent additive white noise. The resulting filter blurs the edges in the image. By using the notion of masking function, an edge preserving filter (EPF) is developed. The EPF is a linear weighted combination of a stationary Wiener filter and an identity filter where the weights are determined using the spatially varying masking function. The usefulness of the algorithm is illustrated using a real image.