Local normalized cross correlation for geo-registration

Abstract

Real time geo-registration is a difficult problem to solve due to issues like partial or complete contrast reversal, partial change in scene and occlusion. In this paper, a local normalized cross correlation based approach is presented that reduces the impact of the above mentioned issues and improves the performance of a geo-registration system. Further enhancements to the proposed approach are also presented to reduce the processing time and improve the performance speed. Extensive experimentation on the proposed approach was conducted to evaluate its performance against standard geo-registration approaches.