Multi-scale image registration based on strength correspondence weighted graph

This paper proposes a novel strength correspondence weighted graph for multi-scale image registration. First, the strength correspondence weighted graph is proposed based on the structure information and attributes relations of graph. Second, a similarity scale parameter obtained which can adjust adaptively scale based on the similarity of the different feature points. Third, the energy is embedded in the strength correspondence weighted graph, which not only considers distribution of the energy but also considers the level of energy, which can better describe the similarity between the features of the image and find the correspondence between the features. Experiments on real-word images demonstrate that the proposed method is effective and accurate for multi-scale image registration.