基于模糊神经网络的彩色图像滤波器研究

为实现彩色图像噪声的滤波，基于模糊神经网络技术构建了一种新型的彩色图像滤波器．该滤波器通过对滤波窗口的彩色图像像素矢量进行模糊加权来判断邻近像素和中心像素的关系，针对不同性质的噪声由神经网络的自学习和自组织功能来自动调节滤波器的权值以实现噪声的滤除．用Flowers图像和Lena图像对经过训练的滤波器和矢量中值滤波器进行了测试对比．结果表明：模糊神经网络彩色滤波器无论对单纯的脉冲噪声及高斯噪声，还是二者的混合噪声，其滤除能力都要优于矢量中值滤波器，并且有较好的边缘和细节保持能力．

Color filter based on fuzzy neural network

A new filter based on Fuzzy Neural Network (FNN) judges the relation of neighbour pixel and center pixel according to a fuzzy weighted averaging operation performed on pixel vectors in the filter window, while the weights are adjusted automatically by self - study and self - organizing functions of the neural network , so the noise component is filtered out. The new FNN filter was compared with vector median filter (VMF) using image Flowers and Lena. Experimental results indicate that FNN filter has better effect on impulse noise, Gaussian noise and mixed noise than VMF, especially on preserving edges and fine image details.