Georgia Institute of Technology College of Sciences

This talk is an introduction to mathematical approaches to image processing: using variational approaches and PDE based method. Various problems and a few different approaches will be introduced.

The classic Pick Interpolation Problem asks: Given points z_1, z_n and w_1, w_n in the unit disk, is there a function f(z) that (1) is holomorphic on the unit disk, (2) satisfies f(z_i)=w_i, and (3) satisfies |f(z)|=1 In 1917, Pick showed that such a function f(z) exists precisely when an associated matrix is positive semidefinite. In this talk, I will translate the Pick problem to the language of Hilbert function spaces and present a more modern proof of the Pick problem. The benefit of this approach is that, as shown by J. Agler in 1989, it generalizes easily to the two-variable setting. At the heart of the proof is a method of representing bounded analytic one and two-variable functions using Hilbert space operators. Time-permitting, I will discuss recent results concerning the structure of such representations for bounded two-variable analytic functions, which is joint work with G. Knese.

Abstract: In this talk, I will use two examples, the influence prediction in social media, and the short path in engineering, to illustrate how we use differential equations to establish models for problems in social science and engineering, and how to use mathematics to design efficient algorithms to compute the solutions. The talk is mainly for first or second year graduate students, and it is based on collaborative work with several faculty members and graduate students in SoM, ECE, CoC.

We present several main models in this area including random polymers. We then explain some open problems big and small as well as a few of our related results.