Differential evolution with Powell´s direction set method in medical image registration

Powell´s direction set method (PDSM) is widely used in image registration. However, its performance is found to be strongly dependent on the initial solution. In this paper we propose a novel combination of PDSM with initialization by the differential evolution genetic algorithm. To further improve the convergence, a multi-resolution approach is taken. Mutual information is chosen as the error metric for this investigation. The results presented show that this proposed method does significantly improve the performance of PDSM and it is efficient and robust.