摘要
本文提出了一种称为特征散度的新概念和一种基于特征散度的图像模糊C-均值(FCM)聚类分割方法。该方法采用特征散度取代传统的欧氏距离作为图像像素与C-聚类典范值之间的差异性度量。本文将新方法与一些传统的FCM算法同时应用于图像分割,并采用形状测度和均匀测度评价各算法的分割性能,结果表明新方法对不同类型的图像都具有良好的分割性能。
In this paper, we introduce a new concept called feature divergence and propose an image segmentation technique using fuzzy c-means (FCM) clustering based on feature divergence. In this new FCM algorithm, traditional Euclidean norm metric is replaced by feature divergence as the measure of dissimilarity between pixel and cluster prototypes. The effectiveness and generality of this new algorithm are evaluated and justfied using uniformity measure and shape measure in the experiments of segmenting various real images along with several existing FCM methods.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1998年第4期462-467,共6页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金
关键词
特征散度
图像分割
FCM
聚类分割
Feature Divergence, Fuzzy C-Means Clustering, Image Segmentantinon
