Color image registration via quaternion Fourier transform and parametric template method

In this paper, we present a new approach that combines quaternion Fourier transform and parametric template method for registering color images that allows for scaling, translation, and rotation. We can accurately estimate model geometrical parameters based on the properties of quaternion Fourier transform. Inspired by parametric template method that allows fast estimation of the model parameters by a direct linear calculation, we conceived a parametric template space by a set of magnitude of quaternion Fourier transforms. Experiments with real-world scenes demonstrate that the proposed method can not only obtain high estimation accuracy in the presence of image translation, rotation and scaling, but also is conceptually simple, easy to implement, and computationally efficient because no iterative steps are involved.