Two-Dimensional Recursive Filtering of Noisy Images with Edge Preserving Properties

Abstract

The filtering problem for noisy non homogeneous images is considered. An edge preserving, two-dimensional, recursive filtering algorithm, closely paralleling the one-dimensional Kalman filter and involving a very modest computational burden is derived. The filter is obtained as a combination of a one-dimensional Kalman predictor and of a one-dimensional fixed-interval smoother. Experimental results show the effectiveness of the new filtering scheme.