Math Dept Logo, Info

Colloquium February 10

Date: Friday February 10

Time: 16:00 to 17:00

Place: 455 McBryde (Commons Room)

Speaker: Misha Kilmer of Tufts Univ

Title: Kronecker Product Approximation for 3D Imaging Applications

Abstract

We derive Kronecker product approximations, via tensor (i.e. multi-way array) decompositions, to the severely ill-conditioned matrices that arise in three-dimensional image processing applications. We show that in a special case, our approximations are optimal. Our primary goal is to use these approximations to construct preconditioners for iterative regularization techniques; the resulting preconditioned algorithms allow us to restore three dimensional images in a computationally efficient manner. Through examples in microscopy and medical imaging, we show that the Kronecker approximation preconditioners provide a powerful tool that can be used to improve efficiency of iterative image restoration algorithms. This is joint work of Misha Kilmer (Tufts), James Nagy (Emory) and Lisa Perrone (Tufts).


Return to