Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Coupling Image Restoration and Segmentation: A Generalized Linear Model Approach

Deformable models minimizing a regularized image energy provide a popular and versatile framework for image segmentation. A wealth of energy functionals, each representing a different model hypothesis, has been proposed. In parallel, algorithms for finding minimizers of these functionals have been developed. We present a new class of image energies derived from a Generalized Linear Model formulation of the photometric properties of the image. This extends energy functionals based on the exponential family by an additional transformation function modeling the imaging process, enabling a principled combination of image denoising, deconvolution, inpainting, and segmentation. With a Total Variation regularizer, an exact convex relaxation of the energy can be found if the transformation is the identity. In all other cases, an upper bound on the convexified energy can be derived and minimized. We present the key theoretical results and demonstrate the capabilities and performance of the framework using a new split-Bregman minimization algorithm on the convexified problem.

Type of Seminar:
IfA BISON Seminar
Prof Ivo F. Sbalzarini
MOSAIC Group, ETH Zurich, Switzerland
Oct 04, 2011   13:30

Contact Person:

Elias August
No downloadable files available.
Biographical Sketch: