在World Wide Web上检索图像和视频信息

提出了一种综合利用图像和视频的文本信息和基于内容的视觉特征进行World Wide Web上图像视频信息检索的原型系统。在这个原型系统中，一个完整的Web上图像和视频信息的处理过程包括：(1)通过探索器从Web上自动收集信息；(2)在文本和视频特征两个域内同时进行分析；(3)图像和视频信息分类；(4)标定索引，进行快速检索。实验结果表明，利用这个系统，可以获得较高的视觉分类率。

Retrieval of Images and Videos on the World Wide Web

In this paper, a prototypical visual information system for searching for images and videos on the World Wide Web is proposed. In the prototype, images and videos are catalogued by combining text-based processing with content-based visual analysis of the images and videos. Besides, a complete processing procedure for images and videos on the web has been studied, including (1) information collecting by automated agents, (2) processing in both text and visual feature domains, (3) cataloging image and video, (4) making index for fast search and retrieval. The experimental result shows that a higher cataloging performance can be achieved by using the prototype.