A remote sensing adapted image registration method based on SIFT and phase congruency

As a reliable and effective feature descriptor, SIFT garners more and more attention in image registration for its capability of scale and rotation invariance. This paper proposes a novel image registration method to utilize SIFT in spectrum disparity cases. The new method uses the local correlation based on phase congruency and is introduced to reinforce original similarity measurement and obtain a more accurate tie point set. With deformation parameters estimated by SIFT, a correlation coefficient is more invariant to scale and rotation by local affine transform. To test the efficiency and effectiveness of this method, three image sets captured from different remote sensing sensors are employed to register a sensed one with a referenced one. Results show that the novel method proposed in this paper achieves higher accuracy and has better registration results than other SIFT-based methods.