Image restoration using fast modified reduced update Kalman filter

Abstract

Some modifications to the reduced update Kalman filter (RUKF) as applied to the filtering of images corrupted by additive noise are proposed. The computational complexity of RUKF is reduced by reducing the state dimensionality. The RUKF is modified using the score-function-based approach to accommodate the non-Gaussian noise. The image is modeled as a nonstationary mean and stationary variance autoregressive Gaussian process. It is shown that the stationary variance assumption is reasonable if the nonstationary mean is computed by means of an edge-and-detail-preserving spatial filter. Such a filter is described