Dissertation Defense

Optical System Identification for Passive Electro-optical Imaging

Joel LeBlanc
3316 EECSMap


A statistical inverse-problem approach is presented for jointly estimating camera blur from aliased data of a known calibration target.  Specifically, a parametric Maximum Likelihood (ML) PSF estimate is derived for characterizing a camera’s optical imperfections through the use of a calibration target in an otherwise loosely controlled environment. The unknown parameters are jointly estimated from data described by a physical forward-imaging model, and this inverse-problem approach allows one to accommodate all of the available sources of information jointly. These sources include knowledge of the forward imaging process, the types and sources of statistical uncertainty, available prior information, and the data itself.  The forward model describes a broad class of imaging systems based on a parameterization with a direct mapping between its parameters and physical imaging phenomena. The imaging perspective, ambient light-levels, target-reflectance, detector gain and offset, quantum-efficiency, and read-noise levels are all treated as nuisance parameters.  The Cram\'{e}r-Rao Bound (CRB) is derived under this joint model, and simulations demonstrate that the proposed estimator achieves near-optimal MSE performance.  Finally, the proposed method is applied to experimental data to validate both the fidelity of the forward-models, as well as to establish the utility of the resulting ML estimates for both system identification and subsequent image restoration.

Chair: Professor Alfred Hero