摘要： 针对基于内容的图片搜索存在召回率低及匹配速度较慢的问题，在词汇树的基础上，利用模糊量化加以解决。把从图像中抽取到的SIFT特征利用词汇树模糊量化到单词中，从而将图片转为用向量表示，同时用向量间的比较测量图片相似度。实验结果表明，该方法可以有效缩短响应时间，提高搜索结果的召回率。

关键词: 图片搜索, 词汇树, 模糊量化

Abstract: Regarding the inadequacy in recall rate and match speed in content-based image search, an effective method based on vocabulary tree is presented, using fuzzy quantification. The SIFT(Scale Invariant Feature Transform) descriptors are extracted from the query image, which fuzzily quantifies these features to words using vocabulary tree. In this way, the query image can be represented as a weight vector. The query vector is compared with database vectors to measure the image similarity. Experimental results show this method can greatly shorten the response time, and improve the recall rate significantly.

Key words: image search, vocabulary tree, fuzzy quantification

中图分类号: 

• TP311.52