Accurate hyperspectral and infrared satellite image registration method using structured topological constraints

As the hyperspectral and infrared characteristics of satellite images are different, the accurate registration between these two kinds of images is a challenge. Considering the robustness of scene structure, this paper proposes an accurate hyperspectral and infrared satellite image registration method using structured topological constraints. First, Delaunay triangulation is used to correlate the scattered feature points, and these local points are combined in a graph arrangement to form triangular mesh. And then, based on the correspondences between local points, triangular edges and triangular surfaces, the multi-level structured topological constraints are established to dynamically remove the outliers. Further, based on the established spatial topology, an image fitting method using multiple transformation models is proposed, which can solve the problem of local distortion of wide-band satellite images rapidly through region merging. Compared with the existing state-of-the-arts, the experimental results show that the proposed method has high registration accuracy, good robustness and fast computational speed for satellite images.