基于可变步长PCNN赋时矩阵高斯噪声滤波

提出了一种用脉冲耦合神经网络(pulse-coupled neural network,PCNN)赋时矩阵定位噪声、分类滤波并能自适应调整灰度补偿步长的高斯噪声滤波方法。使用PCNN求得含噪图像的赋时矩阵,再在滤波窗口中根据目标像素与周围像素的点火时刻关系确定噪声点,对噪声点分4类滤波:增加、减少可变灰度步长,维纳滤波和中值滤波,最后对图像维纳滤波平滑小噪声点。实验结果表明,该算法能有效去除噪声且较好保持图像的边缘细节,与同类高斯滤波方法进行峰值信噪比、信噪比改善因子比较,验证了算法的有效性。

Gaussian noise filter using variable step PCNN time matrix

A novel Gaussian noise filter is put forward,which can identify noise by using pulse-coupled neuron network(PCNN) time matrix,classify noise into four types and filter noise through a adaptive intensity update function.Firstly,time matrix of a noisy image is calculated by using a simplified PCNN.Then noise is classified into four types based on the firing time of the central neuron and the neighboring neurons in the linking window.Four filter methods,which are to increase intensity step,decrease intensity s...