TITLE:

Three topics on signal and images processing

DATE:

Friday, November 7th 2014

TIME:

3:30 PM

LOCATION:

GMCS 214

SPEAKER:

Dr. Jerome Gilles. Department of Mathematics at San Diego State University

ABSTRACT:

As a new faculty at SDSU, the purpose of this talk is to introduced some of my research interests. I will present three different topics:

– Atmospheric turbulence mitigation algorithms: long-range imaging systems are affected by the effect of the atmosphere. Indeed, the presence of turbulence creates local modification of the reflective index resulting in the presence of both blur and geometric distortion on the observed image. I will show how we can model this problem and design specific image processing algorithms to retrieve a corrected image.

-Cartoon+texture decomposition: image segmentation is a very challenging problem, especially in the presence of textures. I will show we can model a certain category of textures and how to separate them from the other “objects” of the image.

-Empirical wavelet: classic wavelets are build independently of the analyzed signal/image which lead certain limitations. I present here a new approach to build adaptive (i.e signal/image dependent) wavelets. Based on the analysis of electroencephalogram signals, I will show that such approach leads to more accurate time-frequency representation than classical wavelets.

HOST:

Dr. Michael O’ Sullivan

DOWNLOAD: