December 2, Wednesday 14:15, Room 303, Jacobs Building

Recent results in geometric modeling and point processing

Lecturer : Andrei Sharf

Lecturer homepage : http://www.idav.ucdavis.edu/~asharf

Affiliation : Center of Visual Computing, Shenzhen Institute of Advanced   
                     Technology(SIAT), Chinese Academy of Sciences, Shenzhen, China


    Most 3D shapes are nowadays acquired using range scanning devices.
    Currently, scanners are capable of capturing complex shapes, large
    urban scenes and lately even motion. The initial representation of
    the shape consists of several properly transformed depth images,
    resulting in a point sampling of the surface. Typically, scan data
    consist of missing parts, noise in point coordinates and orientation,
    outliers and non-uniform sampled regions. Without prior assumptions
    and user interventions, the reconstruction problem is ill posed;
    an infinite number of surfaces pass through or near the data points.
    One of today's principal challenges is the development of robust point
    processing and reconstruction techniques that deal with the inherent
    inconsistencies in the acquired data set.


    In my talk I will present recent advances in geometric modeling,
    processing and reconstruction of point data. I will describe a
    deformable model for watertight manifold reconstruction. The model yields
    a correct topology interpretation of the reconstructed shape and allows
    topology control to a certain extent. Next, I will present a
    topology-aware interactive reconstruction technique. Topological ambiguities
    in the data are automatically detected and user interaction is used to
    consolidate topology reconstruction. Following, I will present a space-time
    technique for the reconstruction of moving and deforming objects. The motion
    of the object is described as an incompressible flow of matter which
    overcomes deficiencies in the acquired data such as persistent occlusions,
    errors and even entirely missing frames. Motivated by recent advancements
    in sparse signal reconstruction, I will present a "lower-than-L2" minimization
    scheme for sparse reconstruction. The sparsity principle gives rise to a
    novel global reconstruction paradigm for sharp point set surfaces which is
    robust to noise.