A quantitative matching technique based on eigenvector approach

Sang Ho Park, Kyoung Mu Lee, Sang Uk Lee

Abstract: In this paper, we propose a new eigenvector-based linear feature matching algorithm, which is invariant to the rotation, translation and scale. First, in order to reduce the number of possible matches, we use a preliminary correspondence test that generates a set of finite candidate models. Secondly, we employ the modal analysis, in which Gaussian weighted proximity matrices are constructed to record the relative distance and angle information between linear features. Then, the modes of the proximity matrices of the two models are compared to yield the dissimilarity measure. Experimental results on synthetic and real images show that the proposed algorithm performs matching of the linear features fast and efficiently and provides the degree of dissimilarity in a quantitative way.