Brief vitae

I received the engineering degree from the Ecole Nationale Supérieure de Physique de Marseille, known now as Ecole Centrale de Marseille, and a specialization in physic, optic and image processing. I started a PhD in image processing in Dec 2003 in the Vista team at Irisa in Rennes under the direction of Patrick Bouthemy and Charles Kervrann. This work has been realized with Inra and the Inria. I defended my PhD in Jan 2007 and since Feb 2007, I'm a Post Doc at Institut Curie in Paris.

Research

Image analysis and image sequence analysis has a growing influence in life science. While in medicine images are used to help the physician to formulate a diagnosis, in biology they can lead to better understand complex mechanisms of life. Images can be provided by several acquisition systems (MRI, X-ray, confocal microscopy). My research activities, focus mainly on the development of statistical approaches for the analysis of time-lapse video-microscopy image sequences. The following table shows the links between applications and the methods.

Non-parametric adaptive estimation

In many problems, one has to face the issue of tuning parameters. For example, bayesian and variationnal approches often require the balance between a regularization term and data term. Moreover, these two framework aims at minimizing a global energy which averages errors. On the contrary, non-parametric adaptive estimation locally adapt the smoothing parameters by minimizing the local quadratic risk of a designed estimator. This risk can be minimized at each point by controlling the tradeoff between the variance and the bias of the estimator.

This approach can be used in a general context for statistical regression. We have developped an estimator for image and image sequence denoising able to preserve spatial discontinuities as well as temporal discontinuities. We have also developped a background substraction procedure for image sequences using this framework. Once, the background substracted, we have shown that objects of interest become much easier to segment. Finally, other problems should also benefit of this framework.

Patch-based approaches

Patch-based approches have shown these last few years its performances for inpainting, denoising and texture synthesis. We have then design an estimator based on the comparison of small patches in an adapted neighborhood for image restoration. The size of the region where these small patches are looked for is given by analysing the bias-variance tradeoff of the estimator as described in the previous paragraph. This approach provides results at the top of the state of the art on traditional benchmark images. Applied to video-microscopy, the very good results will allow to increase the frame rate limited by the signal to noise ration. Finally, extended to color images it should find many applications in other contexts.

On the other hand, the NL-Mean algorithm is also our center of interest. Its striking simplicity and efficiency are very motivating. A more detailled analysis brought us to define a new algorithm, improving its performances. Moreover, since the NL-Mean algorithm is currently analyzed using different frameworks (variational, combination with wavelets, etc..), it is quite interesting to compare them.

Network-tomography

Network-tomography is the name given to the problem of estimating the origine-destination flow on a graph using link counts. It is a typical ill-posed problem. We have proposed to use this approach in order to analyse the intracellular traffic. The intracellular trafic is related to some vesicles transporting some proteins of interest (Rab6) from a donnor compartment to an acceptor compartment. The movements of these vesicles depend on a microtubule network and a very rough analogy can be done with some cars following some roads.

This analogy was our starting point to set up a simulation of the observed trafic in video-microscopy image sequences. The specificity of this approach is that is does not intend to simulate the complex mecanisms explaining the trafic but rather tries to provide image sequences sharing the same statistical properties with the real ones.

The aim of this approach is to be able to estimate from acquired image sequences the parameters of the proposed trafic model in order to caracterize it and to further compare the sequence with other experiments. While being specific to the Rab6 protein, this method should provide a general framework for the analysis of differents proteins involved in the intracellular trafic.

Further...

These selected research topics are also related to the more general and growing bio-imaging context. Video-microscopy plateforms currently produce tera-byte of data. Data access, processing and storage become then very challenging. An other issue is related to the quantitative analysis of these data sets. Moreover, it appears that this can only be done if the context of image sequences is available at the analysis time. These two problems are then finally linked. Some interesting solutions are already suggested by the Open Microscopy Environement team.

People

The work done during my PhD has been supported by the ACI IMPBio as a part of the projet MoDynCell5D (Irisa, Institut Curie, Inra). In this context, i closely collaborated with: Jean Salamero, Jean-Baptiste Sibarita and Victor Racine from Institut Curie; Alain Trubuil and Iklef Béchar from Inra. The other people from the Vista Project involved where Charles Kervrann (Coordinator of MoDynCell5D), Patrick Bouthemy and recently Thierry Pecot.

Finally, here is a (incomplete) list of people with whom i worked or had fruitful discussions: Elise Arnaud, Vincent Auvray, Patrick Bouthemy, Thomas Bréhard, Aurélie Bugeau, Anne Cuzol, Nicolas Gengembre, Patrick Heas, Alexandre Hervieux, Charles Kervrann, Ivan Laptev, Jean-Pierre Le Cadre, Nicolas Papadakis, Gwenaelle Piriou, Jean Salamero, Jean-Baptiste Sibarita, Alain Trubuil, Thomas Veit.

Publications

Here is the list of the publication related to my reasearch.

Thesis

Non-parametric estimation and contributions to image sequence analysis: Modeling, simulation and estimation of the intracellular traffic in video-microscopy image sequences. Université de Rennes 1, Mention Traitement du Signal et des Télécommunications, Jan 2007 (pdf)
Jury:
Head: Michael Unser
Reporters: Philippe Réfrégier, Josiane Zerubia
Examiners: Jean Salamero, Alain Trubuil
Supervisors: Patrick Bouthemy, Charles Kervrann

Journal articles

C. Kervrann, J. Boulanger. Optimal spatial adaptation for patch-based image denoising. IEEE Trans. on Image Processing, 15(10):2866-2878, Oct 2006.

J. Boulanger, Ch. Kervrann, P. Bouthemy. Space-time adaptation for patch based image sequence restoration. IEEE Trans. on Pattern Analysis and Machine Intelligence Jun 2007. (pdf)

Book chapter

J. Boulanger, C. Kervrann, P. Bouthemy. An adaptive statistical method for 4D-fluorescence image sequence denoising with spatio-temporal discontinuities preserving. Biophotonics for Life Sciences and Medicine, F. Faugel, P. Smigielski, A. Brandenburg, J. Fontaine (eds.), pp. 97-113, Fontis Media SA, Lausanne, Suisse, 2006.

International conferences

T. Pecot, J. Boulanger, C. Kervrann, P. Bouthemy. Network tomography for trafficking simulation and analysis in fuorescence microscopy imaging. Proc. IEEE Int. Symp. on Biological Imaging (ISBI'07), Arlington, USA, Apr 2007

J. Boulanger, C. Kervrann, P. Bouthemy. Simulation and estimation of fluorescence in microscopy image sequence. Proc. MICCAI'06 workshop on microscopique image analysis and application to biology, Copenhague, Dannemark, Oct 2006. (pdf)

C. Kervrann, J. Boulanger. Adaptive estimation to variable smoothness for patch-based image denoising. Proc. Annual Meeting of the Institute of Mathematical Statistics, Rio, Brasil, Aug 2006. (invited talk)

J. Boulanger, C. Kervrann, P. Bouthemy. Statistical approach for the analysis of video-microscopy image sequences. Proc. of the International Workshop on Microscopy and Medical Image Processing, Linz, Austria, Mai 2006. (invited talk)

J. Boulanger, C. Kervrann, P. Bouthemy. Adaptive space-time patch-based method for image sequence restoration. Proc. Workshop on Statistical Methods in Multi-Image and Video Processing (SMVP'06), Graz, Austria, May 2006. (Best student paper prize) (pdf)

C. Kervrann, J. Boulanger. Unsupervised patch-based image regularization and representation. In Proc. of European Conf. Comp. Vision (ECCV'06), Graz, Austria, May 2006.

J. Boulanger, C. Kervrann, P. Bouthemy. Adaptive spatio-temporal restoration for 4D fluorescence microscopic imaging. In Proc. of Int. Conf. on Medical Image Computing and Computer Assisted Intervention (MICCAI'05), Palm Springs, USA, Oct 2005.

J. Boulanger, C. Kervrann, P. Bouthemy. An adaptive statistical method for denoising 4D fluorescence image sequences with preservation of spatio-temporal discontinuities. In Proc. Int. Conf. on Image Processing (ICIP'05), Genova, Italy, Sep 2005.

J. Boulanger, C. Kervrann, P. Bouthemy. An adaptive statistical method for 4d-fluorescence image sequence denoising with spatio-temporal discontinuities preserving. In Proc. of Imaging for Medical and Life Sciences (IMVIE2), Ilkirch, France, Mar 2005.

A. Llebaria, J.Boulanger, Y. Boursier. Time domain analysis of polar plumes observed with the LASCO/SOHO coronagraph. ADA-III - Astronomical Data Analysis III Conference, Sant' Agata sui due Golfi, Italy, Apr 2004

National conferences

Ch. Kervrann, J. Boulanger. Adaptation locale optimale pour la régularisation et la représentation d'image à base de motifs locaux. In Proc. of Congrès Francophone de Reconnaissance des Formes et Intelligence Artificielle (RFIA'06), Tours, France, Jan 2006. (pdf)

J. Boulanger, Ch. Kervrann, P. Bouthemy. Approche statistique adaptative pour le filtrage de séquences d'images 3D de fluoroscopie. In Proc of. Congrès Jeunes Chercheurs en Vision par Ordinateur (ORASIS'05), Fournols, France, May 2005.

Research reports

J. Boulanger, Ch. Kervrann, P. Bouthemy. Space-time adaptation for patch-based image sequence restoration. Rapport de recherche IRISA, N°1798, Apr 2006. (pdf)

Ch. Kervrann, J. Boulanger. Local adaptivity to variable smoothness for exemplar-based image denoising and representation. Rapport de recherche INRIA, N°5624, Jul 2005. (pdf)

Misc

Ch. Kervrann, J. Boulanger, P. Bouthemy. Spatio-temporal analysis in 4D video-microscopy. ERCIM News No.60, Special Theme: Biomedical Informatics, Jan 2005.