Neuro-fuzzy filters based on recursive processing and genetic learning

Neuro-fuzzy filters based on genetic learning are a recently introduced class of nonlinear operators that aim at exploiting the powerful paradigms of computational intelligence. These filters adopt fuzzy reasoning to model the noise removal process and then perform an effective noise cancellation without blurring the image details. In this paper, we focus on the latest generation of neuro-fuzzy filters that adopt a multiple-output architecture. These filters are composed of several subnetworks that process different subsets of input data adopting a serial or a parallel approach. Since the filtering action is recursive, even different processing strategies can be combined in the same filtering architecture. As a result, the most appropriate filtering behavior can be learned from a set of training data.