LieMorph: Transformer-based Image Registration Using Flows on Lie Groups


Johannes Bostelmann (Universität zu Lübeck), Jan Lellmann (Universität zu Lübeck)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)

Abstract

We propose a data-driven approach for deformable image registration that extends the TransMorph architecture with a module based on the flow equation formulated on matrix fields. Choosing a suitable Lie group allows us to precisely control components of the deformation that should be ignored by the regularization and provides a power- ful plug-in replacement for the deformation field generator in learning-based methods. Combined with the TransMorph architecture, our method performs favorably on the IXI and OASIS datasets. Our implementation is available at https://github.com/sennhoj321/Liemorph.

Citation

@inproceedings{Bostelmann_2025_BMVC,
author    = {Johannes Bostelmann and Jan Lellmann},
title     = {LieMorph: Transformer-based Image Registration Using Flows on Lie Groups},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_511/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection