Coarse-to-fine registration for infrared and visible images of power grid

We present a novel approach to register the visible and infrared power grid images in this paper. This method is based on the coarse-to-fine structure for infrared and visible registration. First, we obtain the feature points of multi-modality images based on the experience. Second, the similarity geometric transformation model is employed to register the multi-modality images coarsely. At last, fine-grained model is built to correct the deviation of coarse registration. The experiment section shows the good performance of our proposed coarse-to-fine method for visible and infrared power grid images.