|
|
Inverse Problems
| Speaker: |
Mark van Kraaij
|
| Date: |
Wednesday May 23, 2007 |
| Title: |
Total variation
regularization
|
|
Abstract
|
In this talk about inverse problems
we will discuss Total Variation (TV) Regularization. Based on the
Tikhonov regularization we will first recap some of the main results
from regularization by filtering. Then we will focus on variational
regularization methods and zoom in on the TV regularization.
In order to use standard minimization techniques (Steepest Descent,
Newton’s method, etc…) an approximate TV functional is needed. The
necessary gradient and hessian info are derived from this functional.
Examples based on the Fredholm first kind integral equation with
Gaussian kernel are used to illustrate the different methods.
|
|
