Bayesian surface reconstruction

Relatore:  Nicola Moretto - University of Padova
  martedì 30 marzo 2004 alle ore 15.00
The main problem addressed in this talk concerns the reconstruction, from
a limited number of views, of a 3D scene composed of a finite number of
surfaces seen as di®erentiable manifolds. We assign to each surface a geometric
model which describes locally its characteristic (for example it could
be locally compared to a plane), and the reconstruction is based on sampling
an a-priori density which reflects this hypothesis. Measurements are
assumed to be the photo consistency of small patches of the tangent bundle
to the manifold back projected on the cameras. The samples that maximize
the a-posteriori density (obtained fusing together the a-priori and input data
information) are kept and used to update the visibility maps and the reconstruction.
Examples of surface reconstructions by the application of the
proposed method are shown.
The thesis also presents a second project concerning the development of a
new high-speed camera (the ¢-Cam), whose working technique is based on a
back-projection process typical of algorithms used in computer tomography.

Ca' Vignal 3 - Piramide, Piano 0, Sala Verde

Vittorio Murino

Data pubblicazione
10 marzo 2004

Offerta formativa