Overview
This project develops methods from scientific imaging and computer vision for quantitative analysis of cementitious microstructures. It combines 2D images (such as SEM/BSE, µ-XRF, light microscopy) with 3D volumetric data (such as µ-CT and FIB-SEM) within a common geometric framework.
Project goal
The project develops a reproducible workflow for combining complementary imaging methods into a consistent 3D description of cementitious microstructures. µ-CT provides volumetric information, while SEM/BSE, µ-XRF and FIB-SEM provide higher-resolution structural and chemical information. The project brings these data sources into a shared coordinate system and uses them for quantitative analysis of phases, pores and microstructural geometry.
The challenge: single material, multiple imaging modalities
Cementitious materials are difficult to characterize with a single imaging method. Large 3D scans show the global structure but miss fine details. High-resolution SEM and FIB-SEM data show local phase information but cover only small regions. The project addresses this mismatch by registering 2D sections and local high-resolution observations to 3D µ-CT volumes. Such registration is highly challenging due to low resolution of overlap, dimensional and domain information imbalance, local symmetries and pose ambiguities.
Patch Your Matcher accepted to WACV 2026!
The Correspondence-Aware Image-to-Image Translation Unlocks Cross-Modal Matching via Single-Modality Priors. > more
Needles & Haystacks accepted to WACV 2025!
A Dataset and Benchmark for Domain-Agnostic Image-Based Rigid Slice-to-Volume Registration. > more
Principal investigators
- Prof. Dr.-Ing. Horst-Michael Ludwig
Building Materials
F. A. Finger-Institute for Building Materials Science, Bauhaus-Universität Weimar
- Prof. Dr.-Ing. Volker Rodehorst
Computer Vision in Engineering
Faculties of Media and Civil & Environmental Engineering, Bauhaus-Universität Weimar
- Prof. Dr.-Ing. Dipl.-Chem. Andrea Osburg
Construction Chemistry and Polymer Materials
F. A. Finger-Institute for Building Materials Science, Bauhaus-Universität Weimar
Researchers
- Dr. rer. nat. Christiane Rößler
Building Materials
F. A. Finger-Institute for Building Materials Science, Bauhaus-Universität Weimar
- Dipl.-Ing. Florian Kleiner
Building Materials
F. A. Finger-Institute for Building Materials Science, Bauhaus-Universität Weimar
- Anton Frolov, M.Sc.
Computer Vision in Engineering
Faculty of Media, Bauhaus-Universität Weimar
- Stephan Heitmann, M.Sc.
Construction Chemistry and Polymer Materials
F. A. Finger-Institute for Building Materials Science, Bauhaus-Universität Weimar
Publications
- A. Frolov and V. Rodehorst, ‘Patch Your Matcher: Correspondence-Aware Image-to-Image Translation Unlocks Cross-Modal Matching via Single-Modality Priors’, in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pp. 7913–7924, 2026.
- A. Frolov, F. Kleiner, C. Rößler, and V. Rodehorst, ‘Needles & Haystacks: Dataset and Benchmark for Domain-Agnostic Image-Based Rigid Slice-to-Volume Registration’, in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pp. 7081–7091, Feb. 2025.
Project pages
- Needles & Haystacks: https://xaf-cv.github.io/nh-rs2v/
- Patch Your Matcher: https://xaf-cv.github.io/pym/
Code and research artifacts
- NH-RS2V dataset: https://github.com/xaf-cv/NH-RS2V-dataset
- NH-RS2V baselines: https://github.com/xaf-cv/NH-RS2V-baselines
- Patch Your Matcher: https://github.com/xaf-cv/patch-your-matcher
This project is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - project number 518560113.