Accession Number:

ADA409646

Title:

Computational Methods for Atmospheric Optics

Descriptive Note:

Final rept. 1 Mar 1999-31 Aug 2002

Corporate Author:

MONTANA STATE UNIV BOZEMAN DEPT OF MATHEMATICAL SCIENCES

Personal Author(s):

Report Date:

2002-08-01

Pagination or Media Count:

13.0

Abstract:

The development of efficient non-negatively constrained optimization algorithms for image deblurring. This includes a new pre-conditioner based on a. sparse approximation to the blurring operator. The development of efficient pre-conditioners for the joint phase and object estimation problem in phase diversity. These pre-conditioners were based on the Hessian of the quadratic regularization terms. This paper also contains a careful numerical study and comparison of trust region vs. limited memory BFGS methods for the numerical solution to optimization problems arising in phase diversity estimation. Data for this study was obtained from the US Air Force Maui Spate Surveillance Complex in collaboration with Dr. David Tyler. The development of obtained preconditioned conjugate gradient schemes for volume refractive index turbulence estimation These schemes make, efficient use of the layered structure of the atmospheric turbulence profiles. This layered structure gave rise to block-structured matrices. We employed a block analogue of symmetric Gauss-Seidel iteration as our multi-grid smoother.

Subject Categories:

  • Optics

Distribution Statement:

APPROVED FOR PUBLIC RELEASE