Abstract: A novel method，called Single-Linking Pulse-Coupled Neural Network（SL-PCNN），is proposed to filter impulse noise while keeping image details.The SL-PCNN simplifies the related parameters of conventional PCNN and the threshold can be adaptively selected while no iteration is required，which a noisy pixels can be identified by two times of firing process on the original image and the reversed image.Subsequently，the noisy pixels are filtered by a median filter while keeping the fine information-bearing details.The proposed method can adaptively determine the filtering times based on the noise intensity.The method demonstrates better performance compared to conventional impulse noise filters when the noise intensity is no more than 60% and to PCNN based impulse noise filters when the noise intensity is high.Experimental results on visual illustration and subjective indices show the effectiveness of SL-PCNN.

Key words: image filtering, Pulse-Coupled Neural Network（PCNN）, Single-Linking PCNN（SL-PCNN）, impulse noise

摘要： 为有效滤除灰度图像中的脉冲噪声并保留图像的细节信息，提出了单一链接脉冲耦合神经网络（Single-Linking Pulse-Coupled Neural Network，SL-PCNN）模型。SL-PCNN简化了传统的PCNN参数，可自适应选取滤波阈值，SL-PCNN对原图像和反转图像进行两次点火过程即可定位出噪声点而无需进行PCNN循环，然后用中值滤波器滤除噪声。实验结果表明，在噪声强度不大于60%时，SL-PCNN的性能优于经典的脉冲噪声滤波算法;在噪声强度较大时SL-PCNN的性能优于常见的PCNN脉冲噪声滤波算法，主观及客观评价证明该算法的有效性。

关键词: 图像滤波, 脉冲耦合神经网络, 单一链接PCNN, 脉冲噪声