Optimization-Based Terahertz Imaging

A terahertz (THz) imaging modality based on nonlinear optimization that is not limited by weak scatter or low refractive index is presented. Two-dimensional reconstructions of sub-wavelength dielectric cylinders in air are performed through iterative coordinate descent (ICD) optimization in a Bayesian framework using experimental data obtained by illuminating the objects with a THz beam. Strategies for avoiding local minima and to allow for faster convergence and improved image quality are presented and evaluated.