Abstract
The data fidelity term is a key component of shape registration pipelines: computed at every step, its gradient is the vector field that drives a deformed model towards its target. Unfortunately, most classical formulas are at most semi-local: their gradients saturate and stop being informative above some given distance, with appalling consequences on the robustness of shape analysis pipelines.
In this paper, we build on recent theoretical advances on Sinkhorn entropies and divergences [6] to present a unified view of three fidelities between measures that alleviate this problem: the Energy Distance from statistics; the (weighted) Hausdorff distance from computer graphics; the Wasserstein distance from Optimal Transport theory. The \(\varepsilon \)-Hausdorff and \(\varepsilon \)-Sinkhorn divergences are positive fidelities that interpolate between these three quantities, and we implement them through efficient, freely available GPU routines. They should allow the shape analyst to handle large deformations without hassle.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Aspert, N., Santa-Cruz, D., Ebrahimi, T.: Mesh: measuring errors between surfaces using the hausdorff distance. In: Proceedings 2002 IEEE International Conference on Multimedia and Expo, 2002 ICME 2002, vol. 1, pp. 705–708. IEEE (2002)
Charlier, B., Feydy, J., Glaunès, J.: Kernel operations on the GPU, with autodiff, without memory overflows. www.kernel-operations.io, Accessed 15 Oct 2018
Cuturi, M.: Sinkhorn distances: lightspeed computation of optimal transport. In: Advances in neural information processing systems, pp. 2292–2300 (2013)
Fedorov, A., Beichel, R., Kalpathy-Cramer, J., Finet, J., et al.: 3D slicer as an image computing platform for the quantitative imaging network. Magn. Reson. Imaging 30(9), 1323–1341 (2012)
Feydy, J., Charlier, B., Vialard, F.-X., Peyré, G.: Optimal transport for diffeomorphic registration. In: Descoteaux, M., Maier-Hein, L., Franz, A., Jannin, P., Collins, D.L., Duchesne, S. (eds.) MICCAI 2017. LNCS, vol. 10433, pp. 291–299. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-66182-7_34
Feydy, J., Séjourné, T., Vialard, F.X., Amari, S.i., Trouvé, A., Peyré, G.: Interpolating between Optimal Transport and MMD using Sinkhorn Divergences. arXiv preprint arXiv:1810.08278
Genevay, A., Peyré, G., Cuturi, M.: Learning generative models with sinkhorn divergences. In: Storkey, A., Perez-Cruz, F. (eds.) Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics, PMLR, vol. 84, pp. 1608–1617. 09–11 April 2018. Proceedings of Machine Learning Research
Kaltenmark, I., Charlier, B., Charon, N.: A general framework for curve and surface comparison and registration with oriented varifolds. In: Computer Vision and Pattern Recognition (CVPR) (2017)
Lombaert, H., Grady, L., Pennec, X., Ayache, N., Cheriet, F.: Spectral log-demons: diffeomorphic image registration with very large deformations. Int. J. Comput. Vis. 107(3), 254–271 (2014)
Paszke, A., et al.: Automatic differentiation in PyTorch (2017)
Peyré, G., Cuturi, M.: Computational optimal transport. arXiv preprint arXiv:1803.00567 (2018)
Székely, G.J., Rizzo, M.L.: Energy statistics: a class of statistics based on distances. J. Stat. Plann. Infer. 143(8), 1249–1272 (2013)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer Nature Switzerland AG
About this paper
Cite this paper
Feydy, J., Trouvé, A. (2018). Global Divergences Between Measures: From Hausdorff Distance to Optimal Transport. In: Reuter, M., Wachinger, C., Lombaert, H., Paniagua, B., Lüthi, M., Egger, B. (eds) Shape in Medical Imaging. ShapeMI 2018. Lecture Notes in Computer Science(), vol 11167. Springer, Cham. https://doi.org/10.1007/978-3-030-04747-4_10
Download citation
DOI: https://doi.org/10.1007/978-3-030-04747-4_10
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-04746-7
Online ISBN: 978-3-030-04747-4
eBook Packages: Computer ScienceComputer Science (R0)