Electrical Engineering-Systems Dept.

*** SEMINAR ***

Ronen Gal

(M.Sc. student under the supervision of Prof. Nahum Kiryati and Prof. Nir Sochen)

On the subject:

Progress in the Restoration of Image Sequences Degraded by Atmospheric Turbulence

In principle, a still image can be reconstructed from a turbulent video of a static scene using non-rigid registration and deblurring. However, the performance of such algorithms has so far been limited. We compare the use of classical and state of the art deterministic non-rigid registration algorithms for this task. A novel probabilistic registration algorithm is also tested. In addition, we examine several classical and state of the art deblurring algorithms. We used real turbulent videos to evaluate the different algorithms, and showed that when using state of the art deblurring and non-rigid registration algorithms, excellent restoration results are achieved.