期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2012
卷号：12
期号：12
页码：43-47
出版社：International Journal of Computer Science and Network Security
摘要：This paper studies compact global descriptor for visual search using sparse coding. BoF(Bag of Feature)[1] is widely used in visual search application. But BoF has some problem of large memory usage and update inefficiency of new reference image. To overcome this problem, we propose compact global descriptor which needs small amount of memory. For attain better retrieval accuracy, we divide SIFT[2] features into two part according to its statistical property. And sparse coding[3] is applied for aggregation of local descriptors instead of k-means clustering that is used in existing global descriptor algorithm. The evaluation shows that our approach perform better than existing Test Model.
关键词：Compact Descriptor; Aggregation of local descriptor; Sparse coding
