模糊局部信息C-均值聚类算法的修正  

Modified Fuzzy Local Information C-means Algorithm

作　　者：罗维薇[1] 加小红 

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《兰州交通大学学报》2016年第1期25-29,共5页Journal of Lanzhou Jiaotong University

基　　金：国家自然科学基金(61461025);甘肃省自然科学基金(148RJYA011)

摘　　要：针对模糊局部信息C-均值(fuzzy local information C-means,FLICM)聚类算法因其局部空间信息的局限性而导致图像分割结果存在误差的问题,改进FLICM算法的相似度测量因子,并考虑邻域空间距离、灰度信息以及灰度方差对分割效果的影响,提出一种用于图像分割的模糊局部信息C均值的修正算法(WFLICM).实验结果表明,WFLICM能够估算邻域像素的衰减程度,提高图像的分割性能,在抑制噪声的同时更好地保留图像细节,且具有更好的抗噪鲁棒性.FLICM（fuzzy local information C-means）fails to resolve the misclassification problem due to the limitation of local spatial information.In order to solve this problem,a modified FLICM is proposed for image segmentation,which improves the similarity measurement factor by taking into account the effects of spatial distance information,gray level and variance of gray level of neighborhood pixels.The modified algorithm（WFLICM）can accurately estimate the damping extent of neighboring pixels and can suppress noise at large scale while preserving more image details.Experimental results show that the algorithm can improve the performance of image segmentation and has better robustness to noise.

关 键 词：模糊C均值 聚类 图像分割 邻域信息 灰度信息 灰度方差 

分 类 号：TP302.8[自动化与计算机技术—计算机系统结构]