Sparse Representation in Redundant Systems

CSIC Building (#406), Seminar Room 4122.
Directions: home.cscamm.umd.edu/directions

Anisotropic Classes via Level Sets

Dr. Ron DeVore

CSCAMM / Mathematics at University of Maryland

Abstract:   Classical smoothness spaces (Sobolev, Besov, etc.) have been a staple for classifying functional regularity. Unfortunately, they are not well suited for measuring regularity in image processing or nonlinear partial differential equations. For example, they are typically not invariant regularity classes for time dependent nonlinear PDEs. Also, it has long been recognized that they do not provide a very fine classification of images. This talk will introduce ways to measure regularity via level sets and study the compressibility of these new regularity spaces.