Optimal Feature Matching Method using Bayesian Graph Theory

Local feature matching is an essential component of many image and object retrieval algorithms. Feature similarities between object model and scene graph are complemented with a regularization term that measures differences of the relational structure. In this paper, we present a novel approach to the optimal feature matching using new Bayesian graph theory. First, we will discuss properties of various local invariant feature detectors and descriptors for scale, affine transformation and illumination changes. Second, we propose an efficient features corresponding method using local invariant features and new graph matching algorithm. Main theoretical background of our algorithm is that it can be based on the Bayes theorem and an iterative convex successive projection algorithm used to obtain the global optimum solution for feature matching problem. Finally, we have conducted the comparative experiments between proposed method and existing method on various real images. Experimental results show that our method outperforms clearly rather than the existing algorithms about feature correspondence in two images with rotation or scale transformation and illumination changes.