Turbulence strength estimation and super-resolution from an arbitrary set of atmospherically degraded images

In remote sensing, atmospheric turbulence and aerosols limit the image quality. For many practical cases turbulence is shown to be dominant, especially for horizontal close-to-earth imaging in hot environments. In a horizontal long-range imaging it is usually impractical to measure path-averaged refractive index structure constant C_n² (which characterizes the turbulence strength) with conventional equipment. In this paper we propose a method for estimation of C_n² based just on the available recorded turbulence-degraded image sequence. The method exploits the turbulence-induced image "dancing". C_n² is extracted from the estimated image shifts variance. Experimental comparison with C_n² measurements using a scintillometer shows reliable estimation results. We also estimate image motion with sub-pixel accuracy for the purpose of obtaining a high-resolution image by applying a simple super-resolution procedure. Results of super-resolution for real imagery are presented.