基于图像处理的三支聚类

Three-way clustering based on digital image processing

将数字图像处理中模糊锐化算子与三支聚类进行结合,提出了一种基于图像处理的三支聚类算法。该算法通过逆多元二次核函数将数据集的密度量化为灰度值,对数据总体采用模糊与锐化操作,提取锐化后灰度值较高的数据区域,将低密度区域从原始数据中删除。对灰度值较高的数据采用传统的聚类算法得到不同的类簇,然后对每个类簇利用图像模糊算子得到类簇的核心域,锐化算子得到类簇数据边界域,从而获得每个类簇的三支表示。试验采用不同的UCI数据集,通过比较聚类指标Adjusted Rand Index(ARI),Normalized Mutual Information(NMI)和Adjusted Mutual Information(AMI),验证了该聚类算法的有效性。

A three-way clustering algorithm based on image processing is proposed by combining blurring and sharpening operations in digital image processing with three-way clustering.The proposed algorithm quantifies the density of the dataset into gray values through inverse multivariate quadratic kernel function.By blurring and sharpening operations for the university,we delete the samples with low density and obtain samples with high density.Then,different clusters are produced by traditional clustering algorithm for samples with high density.For each cluster,the core region and fringe region are obtained through blurring operation and sharpening operation,respectively.The experiments on different UCI datasets verify that the clustering algorithm is stable and effective by comparing the clustering indexes adjusted rand index(ARI),normalized mutual information(NMI)and adjusted mutual information(AMI).