提出一种基于图的Laplace谱的点模式匹配算法。该方法在每次迭代过程中,利用Laplace矩阵的特征向量和双随机矩阵计算点之间的匹配概率,然后求解已知匹配点之间的TPS(thin plate spline)变换关系,再利用获得的TPS变换参数使待匹配点集相互逼近。随着点集的接近,由Laplace谱分析方法获得的匹配精度也随之提高。对真实和合成数据的实验验证了该方法的有效性。
An algorithm for point correspondences was presented by using Laplacian spectra of graphs. In each iteration, the correspondence probabilities were computed by employing the eigenvectors of the Laplacian matrix and the method of doubly stochastic matrix. Then the TPS (thin plate spline) deformation model was used to estimate the transformation parameters between the matched points. The accuracy of Laplacian spectral correspondence was improved by bringing one point-set closer to the other with the transformation parameters estimated from the current correspondences. Experiments on both real-world and synthetic data demonstrate the effectiveness of the approach.