Abstract: This paper presents an algorithm of point correspondence based on Laplacian spectra of graphs with probabilistic relaxation. Given two feature points sets, it defines Laplacian matrices respectively, analyzes the eigenvalues and eigenvectors of the matrices, and obtains the initial correspondence probabilities. The final matching results are acquired by using the method of probabilistic relaxation. Experimental results show that the method possesses comparatively high accuracy.

Key words: correspondence, Laplacian spectrum, probabilistic relaxation

摘要： 提出了一种结合概率松弛的图的 Laplace 谱的特征点匹配方法。该方法给定了两个特征点集，并分别定义其 Laplace 矩阵，通过分析该矩阵的特征值及特征向量来获得特征点匹配的初始概率。利用概率松弛迭代的方法获得匹配的最终解。实验结果表明，该方法可以获得较高的匹配正确率。

关键词: 匹配, Laplace谱, 概率松弛

CLC Number: 

• TP391