基于聚类的图像视觉内容检索和索引

Image Retrieval and Indexing by Clustering

图像是一种典型的可以大量获取的多媒体数据,对它们进行内容管理具有实际意义.描述在BOIC系统中提出并实现的基于聚类机制的图像视觉内容检索和索引方法.首先给出以视觉特征、空间结构、语义注释等来表示图像内容的模型;然后给出基于该模型的三个检索算法,包括基于视觉感知的颜色检索算法、轮廓检索算法、主色调扩展检索算法;最后给出采用集簇算法对媒体数据进行聚类的索引机制,它建立聚类索引表来缩小查询范围,从而提高了检索效率.

In this paper, the image retrieval and indexing algorithms by clustering, which implemented in BOIC system are presented. The paper describes a content model characterized with visual features, spatial structure and semantic annotations. Then a color retrieval algorithm based on visual perception, a dominant retrieval algorithm and a layout retrieval algorithm are given. All of the searching can be speeded up with indexing by cluttering, which shrinks the range of searching through. The experiment shows a better performance.