Lundi 17 DECEMBRE, 16H, AMPHI JULES FERRY, 29 RUE D’ULM

Detecting Rigid and Deformable Objects in Video Sequences

Pascal Fua
EPFL

Abstract: I will present the real-time methods we have developed for
detecting rigid and deformable surfaces.  They do no require any a
priori pose knowledge and are designed to work both in 2--D and
3--D. Our algorithms start from a set of wide baseline point matches
between an undeformed image of the object and the image in which it is
to be detected. The matches are then used not only to detect but also
to compute a precise mapping from one to the other.

I will discuss our approach to establishing the correspondences, to
dealing with the large number of parameters involved in modeling
deformable surfaces, and to rejecting erroneous matches.  Finally, I
will demonstrate our method's applicability by using it to
realistically modify the texture of a deforming surface and to handle
complex illumination effects.