"[Tools] GLOBALBIOLM : How to standardize and solve inverse problems?" by Dylan Brault

Seminar by Dylan Brault

Abstract

Inverse problems are widely represented in image processing problems. It is a general framework that aims to retrieve the parameters of the acquisition system, or of an object, based on observations. These approaches require however the knowledge of the image formation model also called the direct model. This model is not invertible and sometimes badly conditionned. Thus, optimization techniques are required to solve the problem.

Implementing inverse problems approaches can easilly become a fastidious task. Indeed, since most of the time, the image formation is adjusted and since prior on the solution are needed to invert the model, the implemention should have lots of modularity. For these reasons, the EPFL developped GlobalBioIm an open-source MATLAB library that allows one to easily solve inverse problems.

This toolbox is based on operator-algebra allowing the user to develop inverse problems approaches with « legos bricks ».

In this presentation, this tool will be introduced and applied on a deconvolution example and on digital holography reconstructions. This presentation will be based on [1].

[1] Soubies, E., Soulez, F., McCann, M. T., Pham, T. A., Donati, L., Debarre, T., ... & Unser, M. (2019). Pocket guide to solve inverse problems with GlobalBioIm. Inverse Problems, 35(10), 104006.

This seminar will be held in French.

Lien Cisco Webex :   https://ujmstetienne.webex.com/meet/ImageSCV