| In this dissertation, I present and analyze a set of novel imaging architectures utilizing coded mask designs and devise a set of algorithms for the computational reconstruction of the captured scene as well as the restoration of the reconstructed images.;The commonality behind all presented designs is that they refer to real imaging systems whose hardware is implemented either at Northwestern or at different institutions or entities by our collaborators. All presented imaging architectures employ coded masks and most of them rely on the principles of compressive sensing. Unlike computational photography techniques which focus on extracting information, not available to conventional cameras, our presented designs are mainly driven by cost-reduction. In particular, we aim at achieving the equivalent performance of expensive imaging systems by utilizing lower-cost hardware solutions and effectively combining them with powerful computational optimization schemes.;Specifically, I present our work on a compressive passive millimeter-wave imaging system. Further, I investigate two compressive video architecture variants for visible light imaging. The algorithms used for the reconstruction of the unknown scene are based on Bayesian methods, dictionary learning and deep learning approaches. |
