Nonlinear filtering of noisy images using neuro-fuzzy operators

A neuro-fuzzy approach to nonlinear filtering of noisy images is presented. A new filter is proposed which aims at combining the advantages of neural and fuzzy paradigms. The network structure of the neuro-fuzzy operator implements a particular mechanism based on fuzzy reasoning which specifically addresses noise cancellation and preservation of image details. The learning method is based on the genetic algorithms (GAs) and yields an effective training of the network in presence of data even if highly corrupted by noise. The results of computer simulations show that the neuro-fuzzy filter is very effective in removing impulse noise and is able to outperform a number of methods in the literature