提出了一种自动图像配准方法。该方法首先使用模板匹配获得图像间的粗匹配点,然后利用这些粗匹配点估计图像之间的变换参数,并通过统计这些变换参数的适用情况,选取出一组最佳的图像变换参数,从而实现了图像间的自动配准。实验验证,该方法在两幅图像间存在平移、转动且重叠区域较少时,仍能准确得匹配这两幅图像。
A novel automatic comer-based method for image registration was proposed. The method established the correspondence between comers in two steps. Firstly, template matching was used to obtain some potential matching points. Secondly, referring to these matching points, several projective transformations were calculated. And a statistical scheme was performed to refine the transformation parameters. Several experiments are given to demonstrate that the proposed method is effective even if the region in these two images small overlap.