使用加权图像标注改进Web图像检索

Improved Web image retrieval by weighted image annotations

为了提高Web图像的检索质量,提出了一种融合文本关键字和图像视觉内容的Web图像检索方法.通过改进的图像自动标注模型,将Web图像本身所蕴含的低层视觉特征映射到图像高层语义特征,即图像文本标注;再将词汇相似性计算技术作为语义信息的度量手段,将图像文本标注转换成带有权重的文本标注;利用贝叶斯推理网检索模型内在的多信息融合能力,将带权重的Web图像文本标注特征和Web文档中的文本信息无缝地融合在一起实现Web图像检索.实验结果表明,将Web中的文本关键字和Web图像视觉内容融合起来可在一定程度上提高Web图像检索质量.

A Web image retrieval method was proposed which combines textual terms extracted from Web documents and image contents in order to improve Web image retrieval. Web image contents were translated into image annotations by the improved automatic image annotation model. Then the technology of term similarity measurement, as the metric form of semantic information, was applied to weighting image annotations. These annotations and some terms extracted from Web documents were introduced into Web image retrieval under the framework of Bayesian inference network which has an inherent fusion capability of multiple information sources. Experimental results show that the method improves image retrieval to some extent by combining Web image contents and terms in Web documents.