彩色图像分割的FCM初始化方法  

Initialization approach for fuzzy C-means algorithm for color image segmentation

作　　者：朱征宇[1,2] 王丽敏[1] 

机构地区：[1]重庆大学计算机学院,重庆400030 [2]软件理论与技术重庆市重点实验室,重庆400030

出　　处：《计算机应用研究》2015年第4期1257-1260,共4页Application Research of Computers

基　　金：国家科技支撑计划重点项目(2011BAH25B041)

摘　　要：针对传统模糊C-均值聚类方法所存在的过度依赖初始聚类中心、计算复杂度高等问题,提出一种新的FCM初始化方法。首先,使用维纳滤波分别对图像的R、G、B分量进行预处理,待转换为LAB色彩空间后,通过二次分水岭方法获取图像的封闭区域,并计算各区域的质心;其次,利用自适应无监督的方法对质心进行筛选和合并,将合并结果作为FCM的初始聚类中心;最后,使用FCM方法进行分割。实验结果表明,该方法不仅能够获得较准确的聚类中心,减少了迭代次数和运算时间,而且能够更好地实现图像的准确分割。Aiming at the problems of traditional fuzzy C-means algorithm,which was excessively dependent on the initial cluster centers and that had high computation,this paper proposed a novel FCM initialization approach. Firstly,it respectively preprocessed the R,G and B components of the image by Wiener filter. After transformed the image into LAB color space,it extracted the closed areas of the image by secondary watershed,and then computed the centroid of the areas. After that,it used the self-adaptive unsupervised screen method to merge centroid,and set the final centroid as the initial cluster centers. Finally,it applied FCM algorithm to image segmentation. The experimental results show that the proposed method can not only obtain more accuracy cluster centers,which decreases the iteration times and the computation time,but also make the color image segmentation more accuracy.

关 键 词：图像分割 模糊C-均值聚类 初始化 维纳滤波 二次分水岭 质心筛选与合并 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]