BMFTR Project BeyondVoxels: A paradigm shift in 3D data analysis for computed tomography

In the BMFTR project BeyondVoxels, a team from the University of Passau, Fraunhofer IIS/EZRT and the Helmholtz Centre Hereon is developing a new method for the rapid processing of huge 3D X-ray data. The wavelet-based approach saves time and storage space and runs on standard computers.

High-resolution 3D scans from computed tomography (CT) provide detailed insights into cultural assets, geological structures, material structures, and industrial workpieces, among other things. However, evaluating them requires enormous computing power and storage space. Researchers at the University of Passau, Fraunhofer IIS/EZRT and the Helmholtz Centre Hereon are developing novel methods to overcome this challenge in the BMFTR project BeyondVoxels.

Similar to a pixel, which refers to a picture elemnt in a 2D image, a voxel describes its 3D counterpart and represents the fundamental elements of a CT dataset. Instead of the previous voxel-based processing of the entire data set, BeyondVoxels uses a hierarchical wavelet-based approach. This technique breaks down data step by step into different levels of detail – similar to a digital map that displays more or less detail depending on the zoom level – and only loads the detailed information. This significantly reduces computing time and storage requirements.

In the project, the researchers are integrating all steps of data analysis – from pre-processing and machine-assisted image recognition to interactive visualisation – into a data analysis pipeline. This forms the basis for a software prototype that can display and analyse huge CT data sets in the terabyte range on standard hardware.

"For Fraunhofer, BeyondVoxels offers enormous potential in the processing and visualisation of very large data sets from our high-resolution computed tomography modalities such as Synchrotron, XXL-CT or GiantEye. BeyondVoxels will help us to continue analysing and visualising our data on conventional hardware in the future," says Professor Christoph Heinzl, project spokesperson and expert in imaging techniques at the University of Passau and Fraunhofer EZRT. Professor Tomas Sauer, Chair of Digital Image Processing, and project manager on behalf of the University of Passau, adds: "This project is another example of the value of cooperation between the university and Fraunhofer, because we can turn basic research into relevant application problems and also put the results into practice – all here in Passau. In fact, everything here is based largely on results from the BigPicture project, from which the Fraunhofer Group emerged."

BeyondVoxels' technology is designed to support research and industry. Possible applications include the analysis of components or samples in materials research or geology, and the development and documentation of cultural assets. Thanks to more efficient processing, even larger amounts of data could be evaluated in the future – and still without expensive specialised hardware.

Funding and project participants

The project is receiving funding under the ErUM-Data Action Plan and the funding guidelines for the thematic areas “Software and Algorithms with a Focus on Artificial Intelligence and Machine Learning”, “Research Data Management” and “Federated Digital Infrastructures” for the Study of the Universe and Matter (ErUM) from the Federal Ministry of Research, Technology and Space (BMFTR) for a period of three years. The project leaders at the University of Passau are Professor Tomas Sauer (Head of the FORWISS Institute) and Professor Christoph Heinzl (Chair of Cognitive Sensor Systems and Fraunhofer EZRT). The Helmholtz Centre Hereon in Hamburg is involved as an external institution.

This text was machine-translated from German.

Images:

The illustrative image shows a view through polarised glasses of a stereoscopically rendered 3D computed tomography dataset. Photo: Professor Christoph Heinzl/University of Passau

The second image shows the participants at the kick-off of the BeyondVoxels project on 20 January 2026 at the University of Passau:

• Helmholtz Centre Hereon: Vojtěch Kulvait, Julian Moosmann
• Fraunhofer Society (EZRT, IIS): Prof. Christoph Heinzl, Astrid Hölzl, Thomas Lang
• FORWISS (UP): Alexander Gall, Michael Niedermeier, Dörte Rüweler, Prof. Tomas Sauer, Kathrin Schiermeier

Guests (online): associated partners:

• Humboldt University of Berlin (Johannes Baller)
• European Synchrotron Radiation Facility (ESRF) (Athanasios Papazoglou)
• Technical University of Braunschweig (Marius Milatz, Max Wiebicke)