A fuzzy noise-smoothing filter

The removal of additive noise in digital images is currently a subject of i ntense interest. Recent research has shown that to be effective the filter should adapt itself to the local structure existing in the image: In region s which have no dominant structure the adaptive filter should act as a line ar statistics filter while in regions with a strong or dominant structure t he adaptive filter should act as an ordered statistics filter. In the past the adaptive filters worked in a discontinuous mode by switching between a linear statistic filter (used when the local region has no dominant structu re) and an ordered statistic filter (used when the local region had a stron g structure). In this case, the adaptation mechanism is separate from the n oise smoothing algorithm and is usually ad hoc. In this paper we describe a new fuzzy filter in which the adaptation mechanism is no longer ad hoc: in stead it is integrated into the noise smoothing algorithm. In addition, the adaptation mechanism works in a continuous mode-changing gradually as we m ove from a local region with no dominant structure to a local region with a strong dominant structure. (C) 2000 John Wiley & Sons, Inc.