TRex is a toolbox for tomography reconstruction using proximal algorithms. It is based on the ASTRA Toolbox (version 1.7.1beta) and works under Linux and Windows with Matlab and Python. As of version 1.0, it supports 2D CPU reconstructions.
TRex v1.0 extends ASTRA in many ways:
- Adds the following algorithms [1]:
- BICAV (Block Iterative Component Averaging)
- BSSART (Block Simplified SART)
- OS-SQS (Ordered Subset Separable Quadratic Surrogates)
- Fixes some issues and bugs with release 1.7.1beta of ASTRA
- Adds implementations for the proximal operators of these algorithms for
Gaussian noise (Least Squares) and Poisson noise (Weighted Least Squares):
- ART
- SART
- BICAV
- OS-SQS
- Adds implementation of Linearized Altenating Direciton Method of Multipliers (ADMM) for two data terms
- Least Squares (LS)
- Weighted Least Squares (WLS)
and three priors - Anisotropic Total Variation (ATV)
- Isotropic Total Variaiton (ITV)
- Sum of Absolute Differences (SAD)
[1] Mohamed Aly, Guangming Zang, Wolfgang Heidrich, Peter Wonka. TRex: A Tomography Reconstruction Proximal Framework for Robust Sparse View X-Ray Applications. arXiv preprint (2016).
The TRex Toolbox is open source under the GPLv3 license.
email: mohamed@mohamedaly.info
Copyright: 2016, Visual Computing Center, KAUST, KSA