Large format CTIS in real time: parallelized algorithms and preconditioning initializers
Real time processing of hyperspectral imaging data of the kind acquired by Computed Tomography Imaging Spectroscopy (CTIS) presents some unique challenges in computational power and data storage. The general approach pursued in this work is a direct application of numerical solution methods implemented in the computing cluster environment of Vanderbilt University’s Advanced Computing Cluster Resource (ACCRE). The examination of four reconstruction algorithms, Simultaneous Algebraic Reconstruction, Cimmino Component Averaging, Expectation Maximization-Maximum Likelihood, and Multiplicative Algebraic Reconstruction Technique were explored. The Multiplicative Algebraic Reconstruction Technique proved robust in quality and speed as implemented with the matrix preconditioning method as employed.