A Hybrid Frequency-Spatial Domain Approach to Image Registration for Super-Resolution Reconstruction

Image registration is generally an indispensable part of super-resolution reconstruction. The accuracy of registration results directly affects the quality of super-resolution reconstruction which utilizes the information of aliased images in sub- pixel displacement. In this paper, we propose a hybrid frequency-spatial domain registration approach to address this problem. It effectively combines both the advantages of frequency and spatial domain algorithms and is implemented in three phases in a coarse-to-fine fashion. Our algorithm is then compared with a conventional frequency domain approach. The simulation results show that the proposed algorithm is more precise and robust than the conventional one.