期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2011
卷号：11
期号：5
页码：35-42
出版社：International Journal of Computer Science and Network Security
摘要：Csurka et. al. proposed a bag-of-keypoints approach which represents an image by a histogram of the number of occurrences of local properties at keypoints. In this approach, Scale Invariant Feature Transform (SIFT) descriptors are utilized for generic object recognition. As an alternative to SIFT, a method based on Speeded Up Robust Features (SURF) are reported to give better performance at greater speeds than SIFT. In this paper, we investigate combination way of SIFT or SURF and current various classifies such as Na?ve Bayes, SVM and so on. We also propose a scheme in which a saliency map is utilized for removing irrelevant keypoints. We demonstrate that removing keypoints based on saliency improves classification rate in some situation.
