Abstract
Computing matrix means is becoming more and more important in modern signal processing involving processing of matrix-valued images. In this communication, we define the mean for a set of symmetric positive definite (SPD) matrices with respect to information-theoretic divergences as the unique minimizer of the average divergence, and compare it with the means computed using the Riemannian and Log-Euclidean metrics respectively. For the class of divergences induced by the convexity gap of a matrix functional, we present a fast iterative concave-convex optimization scheme with guaranteed convergence to efficiently approximate those divergence-based means.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
We term this trend maxel imaging. Maxel stands for MAtrix \(\times \) ELement imaging by analogy to pixel (PIcture \(\times \) ELement) and voxel (Volume \(\times \) ELement).
- 2.
- 3.
- 4.
References
Vemuri, B.C., Liu, M., Amari, S., Nielsen, F.: Total Bregman divergence and its applications to DTI analysis. IEEE Trans Med. Imaging 30(2), 475–483 (2011)
Wang, Z., Vemuri, B.C., Chen, Y., Mareci, T.H.: An affine invariant tensor dissimilarity measure and its applications to tensor-valued image segmentation. IEEE Conf. Comp. Vis. Pattern Recognition 1 (2004)
Wang, Z., Vemuri, B.C.: Dti segmentation using an information theoretic tensor dissimilarity measure. IEEE Trans. Med. Imaging 24, 1267–1277 (2005)
Barbaresco, F.: Innovative tools for radar, signal processing based on Cartan’s ge- ometry of SPD matrices and information geometry. In: IEEE International Radar Conference, pp. 1–6, Rome (2008)
Wang, Y., Han, C.: Polsar image segmentation by mean shift clustering in the tensor space. IEEE Geosci. Remote Sens. Lett. 36(6), 798–806 (2010)
Cowin, S.C., Yang, G.: Averaging anisotropic elastic constant data. J. Elast. 46, 151–180 (1997)
Tsuda, K., Rätsch, G., Warmuth, M.K.: Matrix exponentiated gradient updates for on-line learning and bregman projection. J. Mach. Learn. Res. 6, 995–1018 (2005)
Bigün, J., Granlund, G.H., Wiklund, J.: Multidimensional orientation estimation with applications to texture analysis and optical flow. IEEE Trans. Pattern Anal. Mach. Intell. 13(8), 775–790 (1991)
Arsigny, V., Fillard, P., Pennec, X., Ayache, N.: Geometric means in a novel vector space structure on symmetric positive-definite matrices. SIAM J. Matrix Anal. Appl. 29(1), 328–347 (2007)
Moakher, M.: A differential geometric approach to the geometric mean of symmetric positive-definite matrices. SIAM J. Matrix Anal. Appl. 26(3), 735–747 (2005)
Fiori, S.: Learning the fr\(\acute{e}\)chet mean over the manifold of symmetric positive-definite matrices. Cogn. Comput. 1, 279–291 (2009)
Bini, D.A., Meini, B., Poloni, F.: An effective matrix geometric mean satisfying the ando-li-mathias properties. Math. Comput. 79(269), 437–452 (2010)
Bhatia, R., Holbrook, J.: Riemannian geometry and matrix geometric means. Linear Algebra Appl. 413(2–3), 594–618 (2006)
Bini, D., Iannazzo, B.: A note on computing matrix geometric means. Adv. Comput. Math. (2010)
Burbea, J., Rao, C.R.: On the convexity of some divergence measures based on entropy functions. IEEE Trans. Inf. Theory 28(3), 489–495 (1982)
Nielsen, F., Nock, R.: Sided and symmetrized Bregman centroids. IEEE Trans. Inf. Theory 55(6), 2048–2059 (2009)
Yuille, A.L., Rangarajan, A.: The concave–convex procedure. Neural Comput. 15(4), 915–936 (2003)
Latecki, L.J., Lakamper, R., Eckhardt, U.: Shape descriptors for non-rigid shapes with a single closed contour. IEEE Conf. Comput. Vis. Pattern Recogn. 1, 424–429 (2000)
Liu, M., Vemuri, B.C., Amari, S., Nielsen, F.: Total Bregman divergence and its applications to shape retrieval. IEEE Conf. Comput. Vis. Pattern Recogn. (2010)
Acknowledgments
This work was partly supported by the Ecole Polytechnique, Sony Computer Science Laboratories Inc, and the Indo-French Centre for the Promotion of Advanced Research (IFCPAR) to Nielsen, the NIH grant NS066340 to Vemuri, and the University of Florida Allumni Fellowship to Liu. The author warmly thank Professor Rajendra Batia and all participants of the workshop on Matrix Information Geometries for their valuable feedbacks.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2013 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Nielsen, F., Liu, M., Vemuri, B.C. (2013). Jensen Divergence-Based Means of SPD Matrices. In: Nielsen, F., Bhatia, R. (eds) Matrix Information Geometry. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-30232-9_6
Download citation
DOI: https://doi.org/10.1007/978-3-642-30232-9_6
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-30231-2
Online ISBN: 978-3-642-30232-9
eBook Packages: EngineeringEngineering (R0)