Delaunay三角形剖分约束下的图像配准算法

针对一般的特征点匹配只考虑了特征点本身蕴含的信息，容易发生误匹配的问题，给出一种基于Delaunay三角形剖分(Delaunay Triangulation, DT)的图像匹配算法。该算法对参考图和感知图中可能的正确匹配点进行DT剖分，对剖分后的图形进行对比并找出一致的结构，进而得到正确匹配点。实验结果表明该算法可以有效地抑制SURF（Speed Up Robust Feature，SURF）中的错误匹配，增强匹配可靠性，与RANSAC（Random Sample Consensus, RANSAC）法相比，可以大幅缩短计算时间，在基于点的图像配准方面具有较高的应用价值。 

An image registration algorithm with the constraint of Delaunay triangulation

In traditional matching of feature points, only information of the characteristic themselves were considered. In order to overcome the problem of this, an image matching algorithm based on Delaunay triangulation is given. In the proposed algorithm, the candidate feature points are Delaunay triangulated. The graphics after triangulated are compared and the consistent structure is located. The consistent structure is corresponding to correct matching. The proposed algorithm can effectively eliminate wrong match points and enhance reliability. And compared to RANSAC (Random Sample Consensus, RANSAC), the proposed algorithm achieves substantial reduction in computing time. The experiments show that the effectiveness and reliability of the proposed algorithm, which probably be with importance in application.