Files in this item



application/pdfSevinc_Oktem.pdf (8MB)
(no description provided)PDF


Title:Computational imaging and inverse techniques for high-resolution and instantaneous spectral imaging
Author(s):Oktem, Sevinc
Director of Research:Kamalabadi, Farzad
Doctoral Committee Chair(s):Kamalabadi, Farzad; Blahut, Richard E.
Doctoral Committee Member(s):Bresler, Yoram; Davila, Joseph M.
Department / Program:Electrical & Computer Eng
Discipline:Electrical & Computer Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):spectral imaging
imaging spectroscopy
computational imaging
inverse methods
maximum posterior estimation
Bayesian Cramer-Rao bounds
multiframe image deblurring
parameter estimation of superimposed signals
Phase retrieval
dynamic programming
instantaneous (non-scanning) spectral imaging
slitless spectrometer
photon sieve
image formation
diffractive imaging
space remote sensing
solar spectral imaging
Abstract:In this thesis, we develop a class of novel spectral imaging techniques that enable capabilities beyond the reach of conventional methods. Each development is based on computational imaging, which involves distributing the spectral imaging task between a physical and a computational system and then digitally forming images of interest from multiplexed measurements by means of solving an inverse problem. In particular, in the first approach, a nonscanning spectral imaging technique is developed to enable performing spectroscopy over a two-dimensional instantaneous field-of-view. This technique combines a parametric estimation approach with a slitless spectrometer configuration. In the second approach, a spectral imaging technique with an optical device known as a photon sieve is developed to achieve superior spatial and spectral resolutions relative to conventional filter-based spectral imagers. This technique relies on the wavelength-dependent focusing property of the photon sieve, and multiplexed measurements recorded by a photon sieve imaging system with a moving detector. In each of these two techniques, multiplexed measurements are combined with an image formation model and then the resultant inverse problem is solved computationally for image reconstruction. The associated inverse problems, which can be viewed as multiframe image deblurring problems, are formulated in a Bayesian estimation framework to incorporate the additional prior statistical knowledge of the targeted objects. Computationally efficient algorithms are then designed to solve the resulting nonlinear optimization problems. In addition to the development of each technique, Bayesian Cramer-Rao bounds are also obtained to characterize the estimation uncertainties and performance limits, as well as to explore the optimized system design. The effectiveness of the spectral imaging techniques are illustrated for an application in remote sensing of the solar atmosphere. Lastly, the phase retrieval problem, another inverse problem that arises in the photon-sieve imaging setting with coherent illumination, is studied to devise computationally efficient algorithms. As a whole, the developed spectral imaging techniques enable finer spectral information in the form of higher temporal, spatial, and spectral resolutions. This will enhance the unique diagnostic capabilities of conventional spectral imaging systems in applications as diverse as physics, chemistry, biology, medicine, astronomy and remote sensing.
Issue Date:2014-09-16
Rights Information:Copyright 2014 Sevinc Oktem. Portions of this dissertation have been published elsewhere: IEEE copyrighted papers: 1) F. S. Oktem, J. M. Davila, and F. Kamalabadi, “Image formation model for photon sieves,” in IEEE Int. Conf. on Image Processing (ICIP), 2013, pp. 2373–2377. 2) F. S. Oktem, F. Kamalabadi, and J. M. Davila, ``Cramer-Rao bounds and instrument optimization for slitless spectroscopy,'' in IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), 2013, pp. 2169-2173.
Date Available in IDEALS:2014-09-16
Date Deposited:2014-08

This item appears in the following Collection(s)

Item Statistics