Computational imaging

Computational imaging encompasses any indirect imaging techniques where the final result must be computed from a set of measurements, i.e., not direct imaging. The group was formed to develop high-quality X-ray tomography with sub-micron resolution. Working closely with CTLab we have developed many world-first techniques to correct for sample misalignment, X-ray source instabilities, sample motion, and sample manipulator inaccuracies.

We utilise a unique X-ray source trajectory about the sample enabling the only iterative reconstruction of large cone-beam micro-tomography data. The iterative capabilities now enable the group to integrate more physics into the reconstruction process, giving the potential to achieve quantitative tomography, e.g., through dual-energy imaging.

The group is currently working on developing: (i) laminography capabilities to scan large 2D objects with high 3D resolution, (ii) X-ray tomography of dynamic systems with the option to capture a motion-free tomogram of the sample and/or a time-resolved motion-vector map of the sample, (iii) Understanding X-ray scatter and determining a technique to correct for its detrimental effects.

The group has also undertaken some fundamental work in developing a new imaging technique known as "ghost imaging." We have been involved in the first 2D X-ray ghost imaging at a synchrotron, the first neutron ghost imaging and the first x-ray ghost tomography. We continue to explore the advantages of ghost imaging over conventional imaging in the areas of super resolution and dose reduction.

We also explore various applications of machine-learning (including deep-learning) techniques in areas such as: preconditioner optimisation, x-ray scatter estimation and ghost imaging reconstruction.


Kingston, Andrew profile