一个基于语义的图像检索系统:VisEngine

A Semantic-based Image Retrieval System: VisEngine

介绍了一个基于语义的图像检索系统——VisEngine，该系统采用基于图像主要区域的图像分割方法，分别提取图像前景、背景和全局的视觉和抽象语义内容，构造相应的语义模板。接着把这些特征导入到一个面向对象的中间信息结构中，在此基础上进行多种方式的相似性匹配和检索。系统支持多种查询方式，用户交互界面自然友好。实验表明， VisEngine系统能有效地提高首次用户查询的正确率。

To bridge the semantic gap, a semantic-based image retrieval system is proposed in which, prior to retrieval, main regions with plenty semantic information are segmented from an image. Then semantic visual template is created for these regions on the basis of weighted feature combination. These templates are stored in XML files and are used for semantic-based similarity calculation. Further, the associations between semantic concepts and primitive features are learned from user feedback. Experimental results show good performance of the system on the base of 3 400 images.