基于区域特征的图像配准方法研究

Study on Image Registration Method Based on Region Feature

研究一种快速、准确,适用于大角度旋转的图像自动配准方法。首先利用图像边缘点和角点的重复性提取两图像的特征点,有利于提高匹配速度;再利用双向结构相似度最大和三角形相似相结合的方法获得匹配点对,有利于减少误匹配点对;最后利用结构相似度较大的匹配点对求得配准参数,并对图像进行配准。实验结果证明了该算法的正确性和有效性,并且其能够适用于较大缩放倍数和旋转角度图像的配准。

A quick, accurate and automated image registration method applied to large rotation angle is presented. It obtained feature points using the overlaps of edge features and comer features, which can avoid a heavy computing. The maximum of cross-correlation from both direction and principle of triangles similar was used to detect register point pairs, which can reduce the inaccurate matching of feature points. The registration parameter gained from the matching feature points. Experiment results show its validity and accuracy. It also can use to register image with large rotation angle and zoom.