Event Location: NSH 1507
Bio: Umberto Castellani received his Laurea degree in Computer Science from
the University of Verona in 1999. He had been a Visiting Research Fellow
at the Machine Vision Unit of the Edinburgh University, in 2001 by
working with Bob Fisher. In 2003 he received the Dottorato di
Ricerca(PhD) from Dipartimento di Informatica (DI) – University of
Verona, with a thesis titled Image Based Modelling: from sensory data to
3D models. Now, he is Ricercatore (Research Associate) at the same
department. In 2007 he was Invited Professor for two months at LASMEA
(Universitè Blaise Pascal – Clermont Ferrand, France), working with
Adrien Bartoli. His main research interests concern the processing of 3D
data coming from different acquisition systems such as 3D models from 3D
scanners, acoustic images for the vision in underwater environment, and
MRI scans for biomedical applications. The addressed methodologies are
focused on the the intersection among Pattern Recogniton, Computer
Vision and Computer Graphics.

Abstract: In this talk I will introduce two recent activities developed at the
Vision, Image Processing and Sound (VIPS) lab of University of Verona
(Italy). The first one is focused on 3D shape matching. A saliency
measure is introduced to select few sparse interest points on a 3D mesh.
Then, each point is modeled by a Hidden Markov Model which is trained in
an unsupervised way by using contextual 3D neighborhood information,
thus providing a robust and invariant point signature. Therefore,
matching among points of partial views of the same object is performed
by evaluating a pairwise similarity measure among HMMs. The second work
is based on geo-located image categorization in which categories are
formed by clustering geographically proximal images with similar visual
appearance. Moreover, the proposed strategy permits also to deal with
the geo-recognition problem, i.e., to infer the geographical area
depicted by images with no available location information. The method
lies in the wide literature on statistical latent representations, in
particular, the probabilistic Latent Semantic Analysis (pLSA) paradigm
has been extended, introducing a latent aspect which characterizes
peculiar visual features of different geographical zones.