R519, Astronomy-Mathematics Building, NTU
(台灣大學天文數學館 519室)
Monotone and Multigrid Schemes for Solving Optimal Mass Transport Problem arising from Nonrigid Image Registration
Justin W.L. Wan (University of Waterloo)
Abstract
In image registration, the problem is to transform one image to align with another image. One approach is based on the Monge-Kantorovich mass transfer problem. The goal is to find the optimal mapping M which minimizes the Kantorovich-Wasserstein distance. The optimal mapping can be written as M=grad u, where u satisfies the Monge-Ampere equation. In this talk, we will present a monotone discretization scheme for solving the Monge-Ampere equation. Our approach is to reformulate the Monge-Ampere equation as a Hamilton-Jacobi- Bellman (HJB)equation. We will develop a monotone discretization scheme and show that the method is consistent, stable, monotone, and hence convergent to a viscosity solution. We will then present a relaxation scheme which is a very slowly convergent method as a standalone solver but it is very effective for reducing high frequency errors. We will adopt it as a smoother for multigrid and demonstrate its smoothing properties. Finally,numerical results will be presented to illustrate the effectiveness of the method.