Homography estimation using Analytical Fourier-Mellin Transform

The paper presents a registration algorithm based on the Analytical Fourier-Mellin Transform (AFMT). In the Fourier Mellin registration algorithm, the estimation of the geometrical transformation is only done in the Fourier Transform phase domain which is known by containing the shape information. The proposed algorithm will be used to estimate homography since in the vicinity of key points, we can assume that geometrical transformation is simply a similarity. So, we compute AFMT at each key point location and we use the phase correlation to match features. Then, we estimate homography parameters with RANSAC algorithm. An archeological application will illustrate the efficiency of AFMT registration in image mosaicing context by the construction of Punic Panorama.