一种基于SIFT和SUSAN特征的图像匹配方法

针对SIFT方法在角点检测上的不足,提出了一种基于SIFT和SUSAN混合特征匹配的自动匹配算法。算法应用SIFT和SUSAN两种具有互补特性的局部不变特征,利用SIFT方法检测空间极值特征点,利用SUSAN方法检测角点,结合两种特征点位置,利用SIFT方法生成匹配特征向量,最后根据最近邻特征点距离与次近邻特征点距离之比确定初始匹配点对,实现图像的配准。实验表明该算法有效,能够提高图像的自动匹配准确性。

An image matching method based on SIFT and SUSAN features

An image matching method based on mixed SIFT and SUSAN local features was proposed to improve the SIFT′s lack of corner detection.In this method,SIFT features were complemented with SUSAN corner,using SIFT method to detect the space extreme points and SUSAN method to detect the image corners.Matching feature vectors generated by both SIFT features and SUSAN corners.At the end the ratio of the first and second nearest neighbor distance were used to set up the initial correspondences.Experiments show that the method is effective.