Modified PCNN model and its application to mixed-noise removal

Pulse coupled neural networks (PCNN) model is a bionic system. It emulates the behavior of visual cortical neurons of cats and has been extensively applied in image processing. A modified PCNN model was designed in the filter proposed for mixed-noise removal. The filter consists of two stages. The first stage smoothes small-amplitude Gaussian-noise and detects impulse noise or large-amplitude Gaussian-noise by a modified PCNN model, which uses a linear- attenuate threshold function and outputs weighted-averaging intensities of firing pixels, so it is abbreviated as L&A-PCNN. The second stage uses median filter to recover those detected noises. Setting parameters of the PCNN model is critical in designing an ideal filter, so the parameters of L&A-PCNN model are analyzed and adapt to suit the improvement. Simulation experiments show the advantage of the proposed approach.