Fast computation of mutual information in the frequency domain with applications to global multimodal image alignment

Johan Öfverstedt, Joakim Lindblad, Natasa Sladoje

Abstract: Highlights•We propose a novel algorithm for computing mutual information between images, for all discrete displacements.•We compute the cross-mutual information function (CMIF) efficiently in the frequency domain.•The algorithm produces equivalent CMIF maps to a direct method, while being much faster.•We present a global multimodal image alignment method utilizing the proposed fast CMIF computation algorithm.•The alignment method reaches excellent performance on three multimodal benchmark datasets.