基于模糊神经网络的非线性滤波算法
首发时间:2009-05-12
摘要:针对现有的图像椒盐噪声滤除算法缺乏对小于滤波窗口的图像细节与边缘信息的保护问题,本文在分析现有的图像滤波算法基础上,提出一种基于模糊神经网络ANFIS(adaptive neural-fuzzy inference systems,记为ANFIS)的非线性滤波算法。本文算法将滤波过程分为噪声检测和噪声恢复两个阶段。在噪声检测过程中,用自适应中值原理对图像中的噪声点进行初步检测,然后通过局部模糊隶属度函数对检测出的噪声点进行二次判断,有效提高了噪声的准确度。在噪声恢复阶段,利用ANFIS通过自适应神经模糊推理系统对非线性系统的结构和参数进行辨识与自学习,采用混合学习算法,对前向参数和结论参数分别辨识,在提高精度的同时可加快训练收敛的速度。此算法充分利用了图像的局部特征,实验结果表明对椒盐噪声具有很好的细节保护与噪声滤除能力,其效果明显优于现有的其他滤波算法。
For information in English, please click here
Nonlinear Filtering Algorithm Using adaptive neural-fuzzy inference systems
Abstract:The major drawback of recent image filtering algorithms for removing impulse noise is lack of the ability of pre-serving the image details and edges which are smaller than the size of filtering windows, how to take full advantage of local characteristics and details in images t o improve noise removal, detail-preserving and edge-preserving ability of filters? A nonlinear filtering algorithm using ANFIS(adaptive neural-fuzzy inference systems) is proposed. The proposed filter has a two-stage scheme: detecting noise and removing noise. In order to improve accurate rate of noise detection, noise candidates identified with the noise detection algorithm of the adaptive median filter are judged again by local fuzzy membership function, and the structures and parameters of nonlinear system are identified and learned by ANFIS. This net work uses hybrid learning algorithm to identify former parameters and conclusion parameters. The algorithm improves the precision as well as quickens the training speed. The proposed filter achieves a better performance than the other filters mentioned in this paper in the cases of being effective in noise suppression and detail preservation, even when the images are highly corrupted by IN.
Keywords: double noise detector nonlinear filtering ANFIS Salt-and-Pepper noise
基金:
论文图表:
引用
No.3216645839912421****
同行评议
共计0人参与
勘误表
基于模糊神经网络的非线性滤波算法
评论
全部评论0/1000