Thomas Corpetti
This page is no longer maintained
New address
Address: Equipe COSTEL (UMR 6554)
Université Rennes 2 Haute-Bretagne
Campus Villejean - Maison de la Recherche
6, avenue Gaston Berger
35043 Rennes Cedex, France
Brief Vitae
Graduated of INSA de Rennes in Electrical Engineering since June 1999, and holder of the postgraduate certificate STIR (Signal, Telecommunications, Image, Radar) from the University of Rennes I (Images option), I obtain my PhD in "Signal Processing" in July 2002. I actually have a "Temporally Assistant Professor" position at University of Rennes I.
Research topics
The general issue of my PhD works is about fluid motion analysis in image sequences.

As analysis of fluid phenomena has many interesting applications in various domains (meteorology, oceanography, medical imaging, fluid mechanics experiments), we developed, during this Ph.D, two methods dedicated to the analysis of fluid images. The first one consists to estimate the spatial velocity field from two consecutive images of a given sequence. The main issue of the second method we proposed is to extract significant areas of vorticity or divergence, from a given motion field. Let us develop those two methods.

In a first time, to define an appropriate estimation of the optic flow, we were based on the decomposition of a fluid velocity field in terms of laminar, vorticity or divergence components. We also used the "continuity equation", well known from fluid mechanical specialists, as the starting point of our observations, instead of using the classical optical-flow constraint equation, based on a brightness conservation and not well adapted in that context.

This method has been published in IEEE Transactions of Pattern Analysis and Machine Intelligence. You can download this paper here and see a demonstration for details.

My research has then been focused, in a second way, on the detection and the extraction of singular points and their main characteristics, under a given model. The developed method is based on an analytic representation of the motion field. This method has the advantage to be stable, fast, and to extract also others physicals quantities that are of interest from a fluid mechanical point of view. This method will appear soon in the international "Journal of Mathematical Image and Vision" , and a post-script compressed version can be down-load here.

In collaboration with the Cemagref (a french agricultural and environmental engineering research institute), we actually try to validate all proposed methods from a fluid mechanical point of view.

Recent publications

T. Corpetti, Estimation and analysis of dense motion fields for fluid flows - Ph.D. from Université de Rennes 1, 9 July 2002

T. Corpetti, E. Mémin, P. Pérez. Dense Estimation of Fluid flows - IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(3):365-380, 2002.

T. Corpetti, E. Mémin, P. Pérez. Extraction of Singular Points from Dense Motion Fields: an Analytic Approach  Journal of Mathematical Imaging and Vision , 19(3):175-198, 2003.