Invited Talk February 2007

Dr. Adrien Bartoli

Dr. Adrien Bartoli LASMEA laboratory in Clermont-Ferrand, France will be giving an invited talk in the 3D Computer Vision lecture on Tuesday, Feb. 6, 10:30-12:00, in room MI 00.13.009A and a talk at 14:30-15:30 in the seminar room MI 03.13.010 on "Feature-Driven Direct Non-Rigid Image Registration"

Abstract:

Feature-Driven Direct Non-Rigid Image Registration

The registration problem for images of a deforming surface has been well studied. Parametric warps, for instance Free-Form Deformations and Radial Basis Functions such as Thin-Plate Splines, are often estimated using additive Gauss-Newton-like algorithms. The recently proposed compositional framework has been shown to be more efficient, but can not be directly applied to such non-groupwise warps. We bring two contributions. First, we propose a Feature-Driven approach making possible the use of compositional algorithms for most parametric warps such as those mentioned above. This is demonstrated by an Inverse Compositional algorithm for Thin-Plate Spline warps. Second, we propose a piecewise linear learning approach to the local registration problem. Experimental results show that the basin of convergence is enlarged, computational cost reduced and alignment accuracy improved compared to previous methods.

Biography:
Dr. Adrien Bartoli is a permanent CNRS research scientist (Chargé de Recherche) at the LASMEA laboratory in Clermont-Ferrand, France, since October 2004 and a visiting professor at DIKU in Copenhagen, Denmark for 2006-2009. Before that, he was a postdoctoral researcher at the University of Oxford, UK, in the Visual Geometry Group, under the supervisation of Prof. Andrew Zisserman. He did his PhD in the Perception group (ex MOVI), in Grenoble at INRIA, France, under the supervisation of Dr. Peter Sturm and Prof. Radu Horaud. Since September 2006, he is co-leading the ComSee research team. His main research interests are in Structure-From-Motion for rigid and non-rigid scenarios, in the field of Computer Vision.