New image matching method based on spatial region interrelationships

Image matching is a key stage for various computer vision applications. In this paper, we propose a new image matching method based on the visual content and image objects spatial interrelationships. The matching process is done while modeling the input images on a regions' relative positioning graph. Then the two graphs are matched using an algorithm of graph isomorphism. The proposed solution uses a description of the relative positioning relationships between the different regions of an image and it increases the computational efficiency of the graph matching algorithm.