摘要
用传统模糊c均值聚类算法分割图像时,类内数据空间分布离散.针对这一问题,提出一种基于全局空间相似性模糊聚类算法.算法建立全局空间相似性度量标准和全局灰度相似性度量标准,分别计算图像中任意一点与聚类中心点的空间相似性和灰度相似性;通过调整参数来控制两种特征在节点间差异计算中所占的比重,增强了分割结果中类内数据样本空间分布的连续性.分别对3类具有不同特征的图像进行仿真实验,结果表明,与传统FCM算法相比,本文算法分割结果更加精确,更能满足用户的实际需要.
The data in a class are discrete in spatial distribution when segmenting images using a conventional FCM clustering algorithm. For better results of image segmentation, an improved FCM clustering algorithm is proposed. In this algorithm, a global spatial similarity and a global intensity similarity between the pixels of an image and a cluster center are separately calculated and the proportion of the spatial feature to the intensity feature is adjusted according to the requirements, which enhances the continuity of the segmentation results in spatial distribution. Simulation results show the better segmentation accuracy obtained by the proposed algorithm, compared with the conventional FCM clustering algorithm.
出处
《东北大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2012年第2期178-181,共4页
Journal of Northeastern University(Natural Science)
基金
国家自然科学基金资助项目(81000639)
关键词
模糊C均值聚类
图像分割
全局空间相似性度量
全局灰度相似性度量
空间分布
fuzzy c means (FCM) clustering
image segmentation
global spatial similaritymeasure
global intensity similarity measure
spatial distribution