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.
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
关键词
影像特征
聚类
色矩
BTC
image features
clustering
color moments
BTC
