Mathematics Colloquium


Thursday, November 30, 1995, 4:00pm

102 Bradley Hall

Professor Yuying Li

Cornell University

speaks on

What Can Optimization Do for Image Restoration ?

Abstract. Many image restoration methods are formulated as minimizing a convex function with a quadratic equality constraint; different choices of the objective functions lead to different restoration methods. However, since a discrete image restoration problem is typically very large, existing computational algorithms rarely solve the minimization problem originally formulated.

We propose an iterative minimization algorithm for image restoration, using minimization of the total variation as the objective with an inequality quadratic constraint. In particular, we illustrate that different computational schemes for the same formulation can make significant difference in restoration quality.

Computational challenges of the proposed algorithm and the potential role of parallel computing will be discussed.

This is joint work with Tom Coleman, Adrian Mariano and Fadil Santosa

Tea. High tea will be served at 3:30pm in the Lounge.
Emmy's. Certain refreshments will be available at the Emmy's after the talk.
Host. Dennis Healy will be the host, anybody interested in having dinner with the speaker should contact Dennis (Ext. 6-3327).