基于临近像素空间距离的模糊C均值聚类算法  

Fuzzy C means clustering algorithm based on space distance of nearest pixels

作　　者：王军玲[1,2] 周建林[3] 包芳 王士同[1] 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122 [2]江苏省信息融合软件工程技术研发中心,江苏江阴214405 [3]江苏省江阴职业技术学院计算机科学系,江苏江阴214405

出　　处：《计算机工程与设计》2013年第7期2476-2482,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61170122);江苏省自然科学基金项目(BK2011003)

摘　　要：针对传统图像分割算法对不同类型噪声敏感性缺陷的问题,基于临近像素空间距离的模糊C均值聚类算法即SF-CM(fuzzy C means clustering algorithm based on the space distance of the nearest pixels)算法,采用核化的空间距离公式,将点到点之间的距离转化为点到空间的距离,很好的平衡了考察像素点临近像素点的灰度信息与位置信息间的关系,进一步克服了临近像素的位置差异对考察像素影响不同的缺点。通过在合成图像和自然图像上的大量实验并与几个传统算法进行对比,不仅表现出了很强的抗干扰能力,提高了聚类精度,并且很好的保留了原图像边缘等细节信息,体现出了较强的鲁棒性。For the traditional image segmentation algorithms is sensitive to different type noises, the fuzzy C means clustering algorithm based on the space distance of nearest pixels （SFCM） uses a nuclear space distance formula which makes the distance between dot to dot translate into the dot to the space. So that it well balanced gray level and location information of the pixels around the query pixel, overcome the defect brought by the location differences of the pixels near the query pixel. Experiments performed on a large number of synthetic and real-world images show that it not only demonstrate a strong anti-interference ability to the noise, improve the clustering accuracy, hut also retaine the original image edge detail information, demonstrate stronger robustness.

关 键 词：模糊聚类 C均值聚类算法 空间距离 核函数 鲁棒性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]