Image registration among UAV image sequence and Google satellite image under quality mismatch

In this paper, we developed an Unmanned Aerial Vehicle (UAV) image registration system consisting of UAV image to UAV image registration, UAV image to Google satellite image registration, and registration refinement with a normalized variant of mutual information for quality mismatch problem. We show the limitation of the conventional mutual information for quality mismatch and then suggest using a normalized variant of mutual information to refine registration between UAV image and Google satellite image. Experiments carried out on the realistic UAV image sequence and Google satellite image show that the developed system could provide better UAV-to-satellite image registration and could conquer the quality mismatch problem.