An Investigation of Registration Methods on The Images Degraded by Atmospheric Turbulence

Abstract

Image registration is an essential step of deblurring algorithms which is applied multiple images of a scene to minimize the degradations stemming from time-varying blur and geometric deformations. The proceeding step, optimal image fusion is required as well in order to reconstruct the scene representation success of which heavily depends on the registered frames. In this study we deal especially with the images degraded by atmospheric turbulence. We investigate the efficiency of three registration methods; thin-plate spline based non-rigid image registration, demons algorithm based non-rigid registration and Fourier transform based registration. The main thrust of the study is to demonstrate and evaluate the performance of well-known registration algorithms applied on images suffering from atmospheric turbulence.