Expectation maximisation algorithms for terahertz transmission tomography

Abstract

Terahertz (THz) tomography is a recently developed imaging technique allowing 3D inspection of opaque objects. In this paper, we develop an ordered subsets convex algorithm for THz transmission tomography (THz-OSC). Since the reconstruction quality is highly depending on the THz beam energy, we investigate afterwards a multi-energy version of the algorithm in order to provide a more accurate reconstruction of the acquired sample. This multi-energy approach is validated by reconstructing data from tomographic acquisitions measured with a 84/287 GHz transmission scanner. Then we discuss how this dual-energy approach could be able to extract physical properties of acquired samples in addition to improving 3D reconstruction.