一种基于仿射变换的SURF图像配准算法  

Affine-SURF algorithm for image recognition

作　　者：刘欣[1] 李校林[1,2] 谢灿 何鹏[1] 

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065 [2]重庆信科设计有限公司,重庆400065

出　　处：《电子技术应用》2014年第6期130-132,136,共4页Application of Electronic Technique

基　　金：科技型中小企业技术创新基金项目(11C26215113601)

摘　　要：传统的SURF算法对仿射变化较大的图像配准效果差。为此,提出了一种仿射-加速鲁棒性特征(Affine-SURF)的图像配准算法,通过增加经度角和纬度角不变特征引入仿射形变参数来模拟图像在不同角度的变形。实验结果表明,与SIFT、SURF、MSER等配准算法相比,该算法能够获得更多特征匹配对,提高了算法对仿射变化的鲁棒性。Previous SURF algorithm was not ideal for affine image matching, when the viewpoint had been changed with large affine angle. In order to solve the problem, an affine-SURF （Affine-Speeded Up Robust Feature） algorithm for image recognition is proposed. The proposed algorithm increases two invariant feature parameters of longitude and latitude. The method uses affine distor- tion parameters to simulate to deformation of an image at different angles. Experimental results demonstrate that the proposed method outperforms significantly the SIFT, SURF and MSER, especially when images suffer severe affine distortion.

关 键 词：SURF 仿射 图像配准 affine—SURF 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]