【作者】 曹奎； 冯玉才； 袁芳；

【机构】 华中科技大学计算机学院；

【摘要】 <正>1.引言多维索引结构是基于内容的图像检索(Content-Based Image Retrieval,CBIR)研究的一个基本问题。由于图像特征的表示一般均是高维向量,并且图像的距离度量也不仅限于距离,因此相似检索可视为在高维特征空间中寻找与指定点距离最近的一组点的问题。我们注意到图像数据库一般都非常大(十万或百万更多还原

【Abstract】 As distance function employed is metric and computation of distances is assumed to be very expensive in content-based image retrieval,it is required to improve the speed in answering the queries. In this paper,we consider using distance-based index structures for similarity queries on large metric space.Based on the"optimum partition"approach that uses representative points to partition the data space into subregions in a hierarchical manner,we propose a novel distance-based index structure opttree and its variantη-tree to support fast similarity retrieval.We discuss in detail the decisions and the algorithms that led to opt-tree and its variantη-tree,and present the experimental results that compare our index structure with VP-tree.更多还原