Research

Realtime Onset MatchmovingRealtime Onset Matchmoving

Since October 2011, I work as a post-doc at Institut de Recherche en Informatique de Toulouse (IRIT) with Lilian Calvet, Vincent Charvillat, and Pierre Gurdjos of the Vortex team. The goal of this post-doc is to propose new ways of detecting and identifying circular markers in order to provide realtime augmented reality for cinema sets. In particular, I have been working in cases where markers are difficult to handle such as when they appear small or blurred in the images. A patent on this topic is in the process of being registered.

Lagrangian MTVLagrangian Molecular Tracking Velocimetry using Image Registration

I did my first post-doc at Institut de Mécanique des Fluides de Toulouse (IMFT). The goal of this post-doc was to track the deformations of a moving liquid. This work involves direct image registration to which is combined volumetric reconstruction.

A related publication

Simultaneous Image Registration and Monocular Volumetric Reconstruction of a Fluid Flow
F. Brunet, E. Cid, A. Bartoli
British Machine Vision Conference (BMVC), Dundee (UK), 1-11, September 2011
Abstract | bib | pdf

Range Surface FittingRange Surface Fitting

During the first year of my Ph.D., we investigated some algorithms to efficiently fit parameteric surfaces (such as Thin Plate Splines and tensor product B-splines) to range data (acquired with Time-of-Flight cameras). We were particularly interested in fast methods able to automatically select the complexity of the fitted surfaces (trade-off between goodness-of-fit and smoothness).

A related publication

L-Tangent Norm: A Low Computational Cost Criterion for Choosing Regularization Weights and its Use for Range Surface Reconstruction
F. Brunet, A. Bartoli, R. Malgouyres, N. Navab
Proceedings of 3D Data Processing, Visualization and Transmission (3DPVT), Atlanta (Georgia, USA), June 2008
bib | pdf | slides

Image RegistrationImage Registration

Another part of my Ph.D. thesis is concerned with image registration in deformable environments. I deal with several aspects of the standard approaches to image registration:
- Direct image registration: Simultaneous registration and RoI discovery;
- Feature-based image registration: Automatic tuning of the complexity for deformable and parametric warps.
We also proposed a new type of parametric warps designed to cope with perspective effects: the NURBS Warps.

A related publication

NURBS Warps
F. Brunet, A. Bartoli, N. Navab, R. Malgouyres
British Machine Vision Conference (BMVC), London (UK), September 7-10, 2009
bib | pdf | poster

Monocular Reconstruction of Inextensible SurfacesMonocular Reconstruction of Inextensible Surfaces

I worked in collaboration with Richard Hartley on monocular 3D reconstruction of smooth and inextensible surfaces. We proposed new analytic constraints to solve this problem. We also made use of second order cone programming (SOCP).

Source code / download

The source code for this work is available in the download section of this website.

A related publication

Monocular Template-based Reconstruction of Smooth and Inextensible Surfaces
F. Brunet, R. Hartley, A. Bartoli, N. Navab, R. Malgouyres
Tenth Asian Conference on Computer Vision (ACCV), Queenstown (New Zealand), November 2010
bib | pdf | videos

Binomial ConvolutionsDiscrete Geometry

During the second year master internship, I worked on discrete geometry under the supervision of Prof. Rémy Malgouyres. More precisely, we worked on a derivative estimator for discrete functions. We proved that this estimator, based on discrete convolutions and binomial coefficient, converges towards the correct value with an acceptable convergence rate.

A related publication

Binomial Convolutions and Derivatives Estimation from Noisy Discretizations
R. Malgouyres, F. Brunet, S. Fourey
14th IAPR International Conference on Discrete Geometry for Computer Imagery (DGCI), Lyon (France), April 2008
Abstract | PDF | Poster | BibTeX | BibTeXML | RIS | Docbook | Endnote | MS Office