结合语义的Web图像检索系统

INTEGRATING SEMANTICS INTO CONTENT-BASED IMAGE RETRIEVAL FOR WEB

图像不仅是除文本之外应用最广泛的媒体形式,而且常用来表示其他媒体,是一种直观的表示方式。随着网络的发展以及对多媒体信息的需求增加,在Web上进行图像检索成为研究的热点但大部分图像检索系统只能支持文本查询或者仅仅依赖视觉特征。我们设计的结合语义和视觉内容的Web图像检索系统,利用HTML文件特殊的结构,从包含图像的HTML中自动获得图像语义信息。并通过分析图像的视觉特征,将语义与视觉特征巧妙地结合在同一个检索模型中,建立一个多索引的查询模型,支持通过语义和图像例子的多种查询方式。

Image is not only the most widely used media type except text, but also one of the most popular means of representing and retrieving other multimedia information. Image retrieval has become an active research field with the increasing demand in many new practical applications. Most image retrieval systems only support text retrieval or visual feature retrieval, both of which have limitation and advantage in applications. Our content-based image retrieval system for Web pays more attention to automatically recovering meaningful semantic structures from the HTML containing image data. Through the analysis of visual features, of images semantics and visual features are integrated into the same retrieval model, which supports multi-mode of retrieval.