EE Seminar: Restoring an image of a moving object from a turbulence distorted video

~~Noam Zac, 
M.Sc. student under the supervision of Prof. Nahum Kiryati and Prof. Nir Sochen

Monday, June 15, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Restoring an image of a moving object from a turbulence distorted video

Abstract

Restoring a clean image of a static scene from a turbulent video, using de-warping and de-blurring, was suggested in previous works. In this thesis a novel approach for restoring a clean image of a moving object from a turbulent video is introduced. The previously suggested scheme is generalized by using weighting maps in order to focus each block of the algorithm on the object of interest and disregard the background. The dense optic flow method suggested by Brox et al (2004) was generalized by integrating the weighting maps into the functional in order to focus on certain areas. In addition, the functional suggested by Bar et al (2006) is generalized to use the weighting maps to focus the performance of the de-blur algorithm on areas of interest. Both simulated and real turbulent videos are used to evaluate and show the benefits of using the weighting maps at different blocks of the algorithm.