一种基于补偿模糊神经网络的新型图像滤波方法  

New method for image filtering based on compensatory fuzzy neural network

作　　者：吴定会[1] 李真[1] 纪志成[1] 

机构地区：[1]江南大学电气自动化研究所,无锡214122

出　　处：《仪器仪表学报》2008年第7期1470-1474,共5页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60774030);江南大学创新团队发展计划(2007)资助项目

摘　　要：针对图像滤波难题,将补偿模糊神经网络应用于图像滤波中,提出了基于补偿模糊神经网络(compensatory fuzzy neural network)的图像滤波方法。该方法结合了模糊逻辑的推理能力和神经网络的自适应、自学习能力。同时采用具有快速学习的补偿算法,引入补偿模糊神经元,使学习后的网络具有更高的容错性,并弥补了神经网络学习耗时的缺点,提高了滤波效率。模糊运算采用动态的、全局优化运算,使网络更优化,进一步改善了滤波效果。仿真结果表明,该方法对噪声具有很好地滤除作用,与现有其它滤波方法相比,具有明显的效果。To solve the problem of image filtering, a novel method based on Compensatory Fuzzy Neural Network （CFNN） has been proposed. The method combines the reasoning ability of fuzzy logic and the adaptive and self-learn- ing ability of neural network. A compensatory algorithm with fast learning capability is adopted. Moreover, compensa- tory fuzzy neurons are introduced to make the learned network more fault-tolerant, remedy the time-consuming defect of neural network learning, and improve the efficiency of filtering. Dynamic, global optimization computing is employed by the fuzzy computing to make the network optimal and more effective. The simulation results show the validity of this which has a better filtering effect comparing with other existing filtering methods

关 键 词：图像滤波 补偿模糊神经网络 滤波效率 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]