关键词: 脉冲耦合神经网络, 脉冲噪声, 中值滤波, 边缘形态, 邻域结构

Abstract: A new method is proposed for impulse noise denoising which is based on Pulse Coupled Neural Network（PCNN） and the analysis of the neighborhood structure of the noise point. PCNN is used to ignite the noise image to obtain a firing grade image. Noise points are classified according to the difference of their neighborhood structure in the firing grade image. Median Filter is applied to the noise with simply neighborhood structure. To eliminate the noises with complex neighborhood structure, an approach is introduced based on regional membership. The results show the new method can restrains noise effectively and keep the edges detail as well.

Key words: Pulse Coupled Neural Network（PCNN）, pulse noise, median filter, edge continuity, structure of neighborhood