文献类型：期刊文献

中文题名：一种利用边缘直方图压缩的FCM加速算法

英文题名：An Accelerated FCM Algorithm by Reducing the Data in 1D Histogram

作者：刘重晋[1];刘保福[2];屈军亮[1]

第一作者：刘重晋

机构：[1]北京师范大学信息科学与技术学院,北京100875;[2]河南新乡学院物理系,新乡453000

第一机构：北京师范大学信息科学与技术学院,北京100875

年份：2008

卷号：27

期号：8

起止页码：1013-1016

中文期刊名：机械科学与技术

外文期刊名：Mechanical Science and Technology for Aerospace Engineering

收录：CSTPCD;;Scopus;北大核心:【北大核心2004】；CSCD:【CSCD2011_2012】；

语种：中文

中文关键词：图像分割;FCM算法;数据约减;灰度直方图

外文关键词：image segmentation ; FCM ; data reduction; 1 D histogram

摘要：FCM(fuzzy C-means clustering)算法在图像分割中应用较为广泛,但其运算时间开销过大的缺点限制了它的应用。现有的利用一维直方图进行聚合从而压缩样本数量的方法虽然加快了运算速度,但随着压缩的增强分割效果明显变差。本文利用图像的边缘点能够更多地表现图像细节特征的特性,通过利用边缘点灰度值的统计信息作为聚合的特征,使数据压缩相同程度时分割效果更好更稳定,并且在压缩程度的选择上更加灵活,从而更适合数据位数较高的图像分割。而且其运算速度较FCM算法也有几十倍的提高。
Fuzzy C-Means（FCM） clustering algorithm is widely used in image segmentation, but its disadvantage of long run time restricts its application. The existing FCM algorithms that use the 1 D histogram of the image to compress the data can improve the calculating speed, but as the compression ratio becomes bigger the speed improvement effect becomes obviously poor. In this paper, we present an accelerated FCM algorithm that can overcome the above-mentioned shortcomings. Making use of the feature that the image edge can express the details better, we use the statistical information of the edge of an image as the reducing feature so that clustering data can be sharply reduced more effectively and more steadily. In addition, the degree of the reduction can be modified easily, and its calculating speed increases by tens of times compared with the traditional FCM algorithm.