改进的ASIFT算法在图像配准中的应用  

Improved ASIFT Algorithm Applied in Image Matching

作　　者：卢彬 

机构地区：[1]重庆信科设计有限公司,重庆401121

出　　处：《电视技术》2014年第11期211-214,共4页Video Engineering

基　　金：国家科技重大专项资助项目(2009ZX03003-01);科技型中小企业技术创新基金项目(11C26215113601)

摘　　要：SIFT算法具有良好的图像匹配性能,是图像匹配的经典算法之一。针对传统SIFT算法在仿射变换情况下匹配效果不佳、鲁棒性弱的不足,给出一种改进的图像配准方法。首先对图像进行中值滤波处理,去除噪声。然后对图像做仿射变换,提取图像的SIFT特征点进行匹配。最后通过随机抽样一致性方法消除匹配后初始匹配点中的错误匹配点,提高匹配精度。实验结果表明,与SIFT算法相比,提出的方法具有更好的仿射不变性,在平移、旋转、仿射变换情况下均能得到较高的匹配精度,并且有较好的鲁棒性。SIFT is one of the most classical image matching algorithm for its good image matching performance. Aiming at the poor performance and weak robustness in affine transformation of traditional SIFT algorithm, an improved method of image matching is proposed. Firstly, image noise is eliminated through processing the image by median filtering. Secondly, the affine transformation of the image is made, and it can match with SIFT feature extraction of image. Finally, eliminate tile error matching points and improve the matching accuracy. The experimental results show that, compared with SIFT algorithm, tile affine invarianee of the method has better function. In the translation, rotation, affine transformation situation can get higher matching accuracy, and has stronger robustness.

关 键 词：SIFT算法 中值滤波 仿射变换 匹配精度 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]