摘要
针对红外与可见光图像中特征点匹配的难题,提出一种基于形状上下文的特征点邻域边缘描述与匹配算法.首先采用基于曲率尺度空间的角点检测算法进行特征点提取,并将特征点所在曲线的法线作为主方向,避免了图像的旋转代价;然后搜索相同边缘上最近的特征点,通过计算这2个特征点邻域的边缘的像素点分布直方图构造一个120维的特征点描述符,并进行归一化;最后采用最近邻算法实现特征点匹配.实验结果表明,该算法能够有效地实现对红外与可见光图像特征点的精确匹配.
A point matching algorithm based on shape context is proposed to solve the matching problems of feature points in infrared image and visible image. First, feature points are extracted by the curvature scale space corner detector. Second, the normal direction of each feature point on the curve is adopted as the main direction of the point, making the point descriptor rotation invariant. Third, the nearest feature point of each extracted one on the same edge is searched. And then, the histograms of edge pixels of the two key points regions are constructed to construct a 120-dimensional descriptor, and be normalized. Finally, the feature matching is realized via the nearest neighbor algorithm. Experiments show that the proposed algorithm can match feature points in both IR and visible images efficiently and correctly.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2013年第6期857-864,共8页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(60772151
61075025)
