Neurofuzzy selfmade network for image processing based on CNN networks

The Cellular Neural Network (CNN) is very efficient for image processing tasks. However, there are limitations in its processing capabilities because some tasks can not be learned by a single CNN network. In recent years it has been accepted that a set of CNNs connected in parallel can realize more complex image processing tasks than a single CNN. Also recently it has been reported an architecture of neurofuzzy adaptable network (SIMAP) able to construct its structure and membership functions using only input-output data. In this paper is described the way of associating a CNN for every fuzzy rule in the SIMAP network for making complex image processing tasks, which are impossible to do for a unique CNN.