Spectral Correspondence Using Local Similarity Analysis

This paper presents a novel algorithm for point correspondences using graph spectral analysis. Firstly, the correspondence probabilities are computed by using the eigenvectors and eigenvalues of the proximity matrix as well as the method of alternated row and column normalizations. Secondly, local similarity evaluated by shape context is incorporated into our spectral method to refine the results of spectral correspondence via a probabilistic relaxation approach. Experiments on both real-world and synthetic data show that our method possesses comparatively high accuracy.