基于颜色特征利用色矩与BTC法进行影像聚类  

Image Clustering Using Color Moments and BTC Approach Based on Color Features

作　　者：丁启伟[1] 戴晨光[1] 赵博 

机构地区：[1]信息工程大学测绘学院 [2]65014部队

出　　处：《测绘与空间地理信息》2011年第3期162-164,167,共4页Geomatics & Spatial Information Technology

摘　　要：影像聚类是一种对影像数据进行分组的方法,在基于内容的影像检索中,如果能够利用较低层次的可视特征进行高效的影像聚类,将会大大提高影像检索的精度。文章分别利用色矩法与分块截短编码(BTC)方法提取影像颜色特征,然后采用K均值聚类算法来对两种方法进行聚类分析。实验结果表明,分块截短编码(BTC)方法的聚类精度优于色矩法。Image clustering is an approach to group a set of image data into different meaningful categories.The precision of the image retrieving will be much more improved in the content-based image retrieval if using the low-level visual features to cluster the images efficiently.In this paper,color moment and Block Truncation Coding（BTC） were used to extract color features,and K-Means clustering algorithm was conducted to cluster the image data based on the features.The experiment showed that the method of BTC is better than Color Moment in clustering.

关 键 词：影像特征 聚类 色矩 BTC 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]