摘要
提出一种基于模糊神经网络的彩色图像滤波方法.该方法将滤波窗口内的像素矢量作为模糊神经网络的输入,根据像素间的矢量距离进行模糊化,通过模糊推理实现对各个像素加权求均值,得到中心像素的输出.输入的模糊化和模糊推理参数由神经网络的自学习功能自动调整,实现最优的滤波效果.对样本图像的处理结果表明,该滤波方法对不同类型的噪声均有较好的滤波效果.
A new filter based on a fuzzy neural network (FNN) is proposed for color image processing. The input vectors of the filter are mapped into a fuzzy space by a membership function which is related with the distance criterion between the input vectors. And a fuzzy weighted averaging operation is performed on the vectors inside the filter window to replace the noisy vector at the center. The parameters of fuzzy reasoning, fuzzication and defuzzication are stored as neural elements, and are optimized by the training of the neural network automatically. The experimental results show that the filter is good at the detection of the noisy vectors and has a good performance for the filtering of mixing noises.
出处
《控制与决策》
EI
CSCD
北大核心
2004年第1期69-72,共4页
Control and Decision
基金
国家自然科学基金资助项目(39670607).