Argonne National Laboratory

Upcoming Events

Image Reconstruction Advancements for Absorption and Phase Contrast Tomography

XSD Presentation
Mark Anastasio, Washington University
August 29, 2014 12:00PM to 1:00PM
Building 401, Room A1100
In this talk we review recent advancements in image reconstruction methodologies for a variety of X-ray imaging problems. Even with hardware acceleration, the overwhelming majority of available iterative algorithms for 3D image reconstruction that implement non-smooth regularizers remain computationally burdensome and have not been translated for routine use in time-sensitive applications. In this work, two accelerated variants of the fast iterative shrinkage thresholding algorithm (FISTA) for image reconstruction in absorption- and propagation-based X-ray computed tomography (CT) are proposed. The first algorithm seeks to minimizes a penalized least squares (PLS) cost function involving a TV penalty while the second assumes a penalty formed as the sum of a TV penalty and a wavelet-sparsified l1-norm of the object. We also describe an improved image reconstruction approach for grating-based X-ray phase-contrast CT, in which the 2nd order statistical properties of the object property sinograms are exploited, including correlations between them. The problem is formulated within a multi-channel (MC) image reconstruction algorithm framework in which the three object properties are simultaneously estimated. An algorithm based on the proximal point algorithm and the augmented Lagrangian method are proposed to rapidly solve the MC reconstruction problem.