An implementation of edge-preserving filters using layered fuzzy controllers with learning

To achieve an edge-preserving noise reduction for mixed noise, several filters that employ a fuzzy controller have been proposed. The optimization method of the fuzzy controller parameters is one of the problem for these filters that have cascade connected or layered connected fuzzy controllers. In this paper, the method of tuning parameters in these types of the fuzzy controller is developed and is applied to the edge-preserving noise reduction fuzzy filter proposed by Takashima et al. (1993). Then a new edge-preserving noise reduction filter whose parameters can be optimized by learning is obtained. Simulation result shows the resulting filter can sufficiently improve the noise reduction performance of the original filter.