摘要
提出基于信息熵特征选择和信息瓶颈算法的图像聚类算法,首先提取图像的Gabor小波纹理特征和灰度共生矩阵纹理特征,然后采用信息熵特征选择方法进行特征降维;图像聚类方法很多,其中较为典型的k-means聚类算法,但它过分依赖距离函数和聚类中心的选择,采用信息瓶颈算法对图像进行聚类,信息瓶颈算法不需要定义距离函数,它考虑了样本与特征的关系,不仅压缩了样本的信息,同时又考虑保留特征信息。实验结果表明,提出的方法具有良好的聚类效果。
An image clustering algorithm based on information entropy feature selecting and information bottleneck algorithm is proposed, the Gabor wavelet features and gray-level co-occurrence matrix texture features of each image are extracted, and information entropy is used to select feature and reduce the feature dimensionality.A wide variety of approaches are pro- posed for image clustering, among them,k-means is a classic one,because of k-means clustering algorithms is over-reliance on the performance of distance function and cluster centers, information bottleneck algorithm for image clustering is proposed. Information bottleneck algorithm does not require the definition of distance function,which takes into accountthe relation- ship between the characteristics and the sample,it compress the sample information and at the same time retain the character- istics information.The experimental results show that the proposed clustering method has a good clustering performance.
出处
《计算机工程与应用》
CSCD
北大核心
2010年第34期187-190,共4页
Computer Engineering and Applications
基金
广东省科技计划项目工业攻关项目资助课题(No.2007B010200036)
关键词
信息瓶颈算法
Gabor小波纹理
灰度共生矩阵
信息熵
图像聚类
information bottleneck algorithm
Gabor wavelet texture
gray-level co-occurrence matrix
information entropy
image clustering
