Date of Award

Spring 2008

Document Type

Thesis - Restricted

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

First Advisor

Johnson, Michael T.

Second Advisor

Polzin, Jason A.

Third Advisor

Richie, James E.

Abstract

Post processing of magnetic resonance (MR) images generally make use of cubic interpolation to estimate values off of the Cartesian grid, as it does not suffer from the artifacts caused by linear and nearest neighbor interpolation yet is not as computationally expensive as sine interpolation. In this thesis, we look to investigate whether recent advances in generalized interpolation techniques can be extended to improve MR image quality, while keeping in mind the relative computation cost of competing techniques. We start with head to head comparisons of interpolation techniques on compounded rotations of controlled, simulated data. Next, the impact on phantom and in-vivo data sets when this higher order interpolant is used along with complex data in image magnification is demonstrated. Finally, the non-rigid transformation mapping of MR gradient distortion correction is considered and how this separable one-dimensional technique from the prior chapters and references behaves when extended to a nonseparable application.

Share

COinS

Restricted Access Item

Having trouble?