A fast algorithm for image registration
Jérôme Fehrenbach, Mohamed Masmoudi · Comptes Rendus Mathématique · 2008
An algorithm is presented here to estimate a smooth motion at a high frame rate. It is derived from the non-linear constant brightness assumption. A hierarchical approach reduces the dimension of the space of admissible displacements, hence the number of unknown parameters is small compared to the size of the data. The optimal displacement is estimated by a Gauss–Newton method, and the matrix required at each step is assembled rapidly using a finite-element method.