一种改进的Harris角点检测的图像配准方法

针对在传统的Harris角点检测过程中,手动输入单个阈值可能出现角点聚簇、伪角点等现象,提出了一种改进的Harris角点检测方法的图像配准方法。首先,将图像分割成3×3个无重叠子图,根据每个子图的对比度的大小,来设置每个子图的阈值。然后,采用NCC算法对检测出的角点进行粗匹配。最后,采用RANSAC算法对粗匹配中误匹配点对进行剔除。实验表明:该算法使得检测的角点分布比较均匀,并在图像配准中有效地增加图像匹配点对数,具有良好的实用性。

Image registration method based on improved Harris corner detection algorithm

In traditional Harris corner detection process,manual input of single threshold may lead to corner cluster,false corners and so on.To solve this problem,an image registration method based on improved Harris corner detection algorithm was proposed.Firstly,the image was divided into 3×3 sub images.According to the contrast of each sub image,the threshold value of each sub image was set.Then,the detected Harris corners were roughly matched by NCC algorithm.Finally,the error matching corners in rough matching were removed by RANSAC algorithm.The experimental results show that corner distribution of the algorithm is more uniform,and the algorithm can effectively increase the number of image matching points in image registration and has a good practicability.