本文针对图像拼接方法中的特征点匹配和变换参数求解问题,提出了一种基于最小生成树和TPS变换模型的图像拼接算法。该算法在每次迭代过程中,利用最小生成树的Laplace矩阵获取待拼接图像中特征点的匹配关系,然后估算待拼接图像之间的TPS(thinplatespline)变换参数,再利用这些参数使特征点集相互逼近,最终获得匹配关系和精确的TPS变换参数,实现图像的拼接。实验结果验证了该算法的有效性。
Aiming at feature point matching and transformation parameter solution problems, an image mosaicking algorithm based on minimum spanning tree and TPS transformation model is presented. In each iteration, the algorithm uses Laplace matrices of the minimum spanning tree to obtain the feature point matching of the two images to be mosaicked. Then the TPS (Thin Plate Spline) transformation parameters of the two images to be mosaicked are estimated and used to make the feature point sets closer each other. Finally, the image matching and accurate TPS parameters are obtained, and the images are mosaicked. Experiment results demonstrate the effectiveness of the algorithm.