elsa - an elegant framework for tomographic reconstruction

elsa is an operator- and optimization-oriented framework for tomographic reconstruction, with a focus on iterative reconstruction algorithms. It is usable from Python and C++.

By design, elsa provides a flexible description of multiple imaging modalities. The current focus is X-ray based computed tomography (CT) modalities such as attenuation X-ray CT, phase-contrast X-ray CT based on grating interferometry and (anisotropic) Dark-field X-ray CT. Other imaging modalities can be supported easily and can leverage our extensive suite of optimization algorithms.

CUDA implementations for the computationally expensive forward models, which simulate the physical measurement process of the imaging modality, are available in elsa.

The framework is mostly developed by the Computational Imaging and Inverse Problems (CIIP) group at the Technical University of Munich. For more info about our research checkout our at https://ciip.cit.tum.de/.

The source code of elsa is hosted at https://gitlab.com/tum-ciip/elsa. It is available under the Apache 2 open source license.

Check the readme at the source repository for current build and installation instructions. A good point to get started are our guides, which can be found here. Another good starting point, is the example folder of the repository. There many different scenarios are covered and well documented. The C++ API reference is found here.