基于特征点的对数极坐标变换图像配准算法  

Log-polar transformation based on feature points for image registration

作　　者：雷凯[1] 刘艳滢[1] 王延杰[1] 邢忠宝[1] 尹立敏[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所

出　　处：《光学技术》2007年第3期435-437,共3页Optical Technique

摘　　要：介绍了一种新的图像配准算法,可以很好地解决图像配准中的平移、旋转和缩放问题。算法的实现是首先使用SUSAN算子提取两幅图像的特征角点,剔除虚假的特征点(噪声点),然后使用改进对数极坐标变换和投影相关匹配算法实现特征点匹配;结合亚像素定位技术,可以进一步提高算法的精度。A new algorithm about image registration was presented, which solved translations, rotations and scale transformations problems in image registration process well. The SUSAN algorithm was used to obtain the feature points （interesting points） of the two images. The bad-feature points （the noise points） were eliminated. The improved log-polar transformation and projection correlation matching algorithm were used to acquire the registration parameter of the two images. The sub-pixel location technique was used to improve the registration precision.

关 键 词：图像配准 SUSAN算法 改进对数极坐标变换 亚像素定位 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]