TITLE:

VARIATIONAL IMAGE PROCESSING TECHNIQUES FOR SHAPE AND SURFACE RECONSTRUCTION.

DATE:

Friday, March 16th, 2012

TIME:

3:30 PM

LOCATION:

GMCS 214

SPEAKER:

Frederick Park.
Mathematics Department.
University of California at Irvine.

ABSTRACT:

Variants of the popular and well studied Rudin-Osher-Fatemi (ROF) and
celebrated Mumford-Shah (MS) models continue to have new and interesting
applications. In this talk, we will present two such applications to shape
prior segmentation as well as surface reconstruction from point cloud
data. In the first part of the talk, we will give a brief introduction to
image processing and present the ROF and MS models. We will then discuss
the two variants used to tackle the aforementioned problems. The first is
a polygonal piecewise constant version of the Mumford-Shah model that
incorporates shape priors. The second is a convex formulation based on the
ROF and MS models that utilizes a novel edge detector and initial ‘pseudo’
image for reconstructing surfaces from unorganized data points in space.
Some ongoing and future work related to these topics will also be
discussed.

HOST:

Dr. Peter Blomgren.

DOWNLOAD:

PDF File