A novel technique for edge-preserving smoothing is developed bq, using
median mean neural hybrid filters. This filter structure is represent
ed by the cascade connection of a median filter, a mean filter and a n
eural network. This filter can adapt itself to the various noise envir
onment through the learning of a training image. In the case Ir he, e
a priori data such a training image is unavailable, this filter, can b
e efficiently applied to edge-preserving smoothing for the images degr
aded by the Gaussian and impulsive noises. Moreover, the structure of
the proposed filter is very simple. Finally, the simulation examples s
how the effectiveness of the proposed filters. (C) 1998 The Franklin I
nstitute. Published by Elsevier Science Ltd.