Scale invariant image matching using triplewise constraint and weighted voting

Due to limited computational resource, image matching on mobile phone places great demand on efficiency and scale invariant. Though spectral matching (SM) with pairwisely geometric constraints is widely used in matching, it is not efficient and scale invariant for applications in mobile phones. The main factor that limits its efficiency is that it requires to eign-decomposition of a large affinity matrix when the number of candidate correspondences is large. It lacks scale invariance because the pairwise constraints cannot hold when large scale variation occurs. In this paper, we attempt to tackle these problems. In the proposed method, each candidate correspondence is considered as a voter and a candidate as well. As a voter it gives voting scores to other candidates and also votes itself. Based on the voting scores, the optimal correspondences are computed by simple addition operations and ranking operations, which results in high efficiency. To make the proposed method scale invariant, we propose a novel triple-wisely geometric constraint formed by three potential correspondences with one being the candidate and the other two being voters. The three correspondences constitute a pair of triangles. The similarity of the two triangles is the core of the triple-wisely constraint, which is robust to scale variation. The information of triple-wise constraints are encoded in a 3-dimensional matrix from which the optimal correspondence can be obtained by simple summation and ranking operations. Experimental results on real-data show the effectiveness and efficiency of the proposed method.