Survey of estimation techniques in image restoration

Abstract
Blurred and noisy images can often be represented as nonstationary 2D stochastic processes that can be modeled by a set of linear space-varying state equations, or by an ARMA input-output equation with space-varying coefficients. Liner difference equation models for characterizing both images and their degraded observations are reviewed. The models are then expressed in state-space form suitable for Kalman filtering and in input-output equation form suitable for maximum likelihood parameter identification and ARMA smoothing. Recent methods for blur identification, image parameter identification, and simultaneous image and blur identification are reviewed. The fundamentals of image restoration are briefly summarized, and three approaches are discussed: iterative deterministic regularized restoration, restoration using optimal filtering, and adaptive restoration. Some representative results are given, and recommendations for future research topics are made.

This publication has 35 references indexed in Scilit: