A Homotopy Method in Regularization of Total Variation Denoising Problems
Luis Melara
Colorado College, Department of Mathematics and Computer Science
Tuesday, August 30, 2005 15:00-16:00,
NIST North (820), Room 145
Gaithersburg
Tuesday, August 30, 2005 13:00-14:00,
Room 4511
Boulder
Abstract:
Homotopy methods can often be used to make mathematical programming problems easier to solve.
One source of notoriously difficult problems is total variation image denoising.
In this talk, a direct relationship is established between the radius of the Kantorovich ball guaranteeing the convergence of Newton's method
and the regularization parameter.
This radius of convergence increases with the choice of the regularization parameter.
A homotopy method in the regularization parameter is employed to improve numerical performance of Newton's method applied to these problems.
The talk will summarize a convergence analysis and present numerical results.
Speaker Bio:
Luis Melara received his PhD in Computational and Applied Mathematics from Rice University in 2003.
He held an NRC postdoctoral position at NIST from 2002-2004 under the supervision of Tony Kearsley.
Luis joined the faculty of Colorado College in Fall 2004 as an Assistant Professor in the Department of Mathematics and Computer Science.
Presentation Slides: PDF
Contact: P. M. KetchamNote: Visitors from outside NIST must contact
Robin Bickel; (301) 975-3668;
at least 24 hours in advance.
|
