Fyzikální ústav Akademie věd ČR

Blind Deconvolution in Image Processing

Seminář Úterý, 09.04.2013 10:00 - 11:00

Přednášející: Ing. Filip Šroubek, Ph.D. (Ústav teorie informace a automatizace AV CR, v.v.i.)
Místo: Fyzikální ústav AV ČR, v. v. i., Cukrovarnická 10, Praha 6/ knihovna, budova A, 1.p.
Jazyk: anglicky
Pořadatelé: Oddělení tenkých vrstev a nanostruktur

I will address the problem of estimating a convolution kernel from degraded measurements and removal of the degradation using state-of-the-artalgorithms. This blind deconvolution problem is ill-posed and conclusions will be drawn that additional prior information is absolutely essential for reaching stability. I then provide details on different strategies that secure the aforementioned prior information targeting primarily applications in image processing. In many practical situations the convolution kernel is space-variant, which renders the problem even more challenging. I will briefly survey approximation techniques of the space-variant degradation operator and show results on real data.

Přílohy:

Copyright © 2008-2014, Fyzikální ústav AV ČR, v. v. i.