Meteorological Image Registration Algorithm

Image registration is the process of finding a transformation between two or multiple images gotten at the same or different time through the same or different sensors. Image registration is widely used in navigation and orientation, target identification, sports analysis, three-dimensional visual, data fusion and so on. As one of image registrations, the meteorological image registration provides services in many areas such as weather forecasts, atmospheric researches, all kinds of disaster monitoring and preventions.In this thesis, rigid and non-rigid deformations exist at the same time in the meteorological images gotten at different time. The thesis reviews different methods in image registration and analyzes some of the rigid and non-rigid image registration methods detailedly.The global rigid deformation is modeled by affine transformation. Because of the time efficiency, the thesis chooses feature-based approach to correct the rigid deformation, preparing for non-rigid registration followed.The non-rigid image deformation often appears when the camera angle changes and the cloud remove. The thesis chooses two methods to correct the non-rigid deformation. One is intensity-based algorithm which models the transformation as locally affine but globally smooth. According to the deformation characteristics of the meteorological images and the requirements of the project, the thesis simplifies the original model of the registration, which effectively shortens the registration time. The other uses free-form deformation based on B-Spline to describe the non-rigid deformation. This thesis proves the feasibility of the methods through the evaluating coefficients: "Mutual Information", "Correlative Coefficient" and run-time parameters with a series of experiments.