Object-Based Image Formation in Cluttered Environments from Polychromatic X-ray Data
Spectral computed tomography (CT) has become possible with the development of photon counting X-ray detector technology. Energy selective measurement capabilities of these devices open the doors to many exciting directions in CT research. In this work we assume perfect energy resolution at detectors, which results in a family of monochromatic CT problems. We propose a tensor based iterative algorithm that simultaneously reconstructs the X-ray attenuation distribution for each energy level. Specifically, we model the multi-spectral unknown as a 3rd order tensor where first two dimensions are in space and the 3rd dimension is in energy. This approach allows the design of a regularizer based on low rank assumptions on the multi-spectral unknown where we apply tensor spectral norm penalties. In addition, when accompanied to total variation (TV) it enhances the regularization capability and provides superior reconstructions. Additionally, we have developed an adaptively weighted L2 norm regularizer with excellent edge preserving capabilities. The problem is cast as a convex optimization problem that is solved using the alternating direction method of multipliers (ADMM). Simulation results show that the proposed regularizer is applicable to the spectral CT problem and reliable in recovering multi-linear structures in an inverse problem set up.
Our work introduces a novel polychromatic dual energy imaging algorithm with an emphasis on detection of explosivesF1-A2/F3-A3 Project Overview
Students Currently Involved in Project
- Oguz Semerici