Model-Based Image Reconstruction For THz Imaging Systems.

Abstract

Broadband THz imaging is an emerging technology with a plethora of promising applications in biological imaging, nondestructive testing, security screening, and process control. Despite the inherent potential of THz technology, there are many factors that severely constrain THz systems from large-scale market introduction. Most notable among these factors is the low power of THz emitters and the measurement speeds of THz imaging systems. To live up to its enormous inherent potential, THz imaging has to dramatically increase its acquisition speed without compromising the signal to noise ratio (SNR) of the imaging system. This thesis presents research in the area of inverse THz imaging architectures via model-based image reconstruction. In inverse or indirect imaging, one can use a single point measurement system with a very high SNR to collect a fraction of the measurements needed in direct imaging modalities in order to reconstruct the object. This thesis begins by building on the initial work of time-reversal THz imaging, a simple inverse algorithm, by adapting a waveguide approach first pioneered in ultrasound to effectively increase the numerical aperture of the THz imaging system without compromising the acquisition speed of the system. The waveguide approach was a success because of the 2.6 x improvement in intensity and the approximate 30% improvement in resolution while maintaining the same acquisition time. The second part of the thesis presents the theoretical framework of model-based image reconstruction in the context of THz imaging. Sophisticated algorithms based on this paradigm are developed for image reconstruction in both transmission-mode and reflection-mode THz systems and for both dielectric and metallic objects. The reconstructed images via the model-based algorithms are shown to be significantly better both quantitatively and qualitatively than those obtained via the time-reversal technique.