Simultaneous reconstruction and separation in a spectral CT frameworkPP

We propose a framework to simultaneously separate and reconstruct the physical components of an object observed in spectral Computed Tomography (CT) with hybrid pixel detectors, a technological breakthrough. We encompass the underlying polychromatic model of the X-ray beam with non-differentiable and low-rank priors on the components of the object. We solve a non-convex ill-posed inverse problem with a specific variable metric Forward-Backward algorithm with high speed convergence rate and show results on real data.

This is poster number 16 in Poster Session

Yannick Boursier (Aix-Marseille University, Computer Science & Engineering)
Sandrine Anthoine (Aix Marseille Univ, CNRS, Centrale Marseille, I2M)
Souhil Tairi (Aix Marseille Univ, CNRS/IN2P3, CPPM)
Christian Morel (Aix Marseille Univ, CNRS/IN2P3, CPPM)
computed tomography, image reconstruction, inverse problems, nonlinear optimization