Dictionary Learning and Sparse Coding on Riemannian Manifolds with Applications to Computer Vision and Medical Imaging

Lecture / Panel
For NYU Community

Speaker: Baba C. Vemuri, University of Florida

Existing dictionary learning algorithms rely heavily on the assumption that the data points are vectors in some Euclidean space Rd, and the dictionary is learned from the input data using only the vector space structure of Rd. However, in many Computer Vision and Medical Imaging applications, features and data points often belong to some known Riemannian manifold with its intrinsic metric structure that is potentially important and critical to the application. The extrinsic viewpoint of existing dictionary learning methods becomes inappropriate and inadequate if the intrinsic geometric structure is required to be incorporated in the model. In this talk, I will present a very general dictionary learning framework for data lying on known Riemannian manifolds. Using the local linear structures furnished by the Riemannian geometry, a novel dictionary learning algorithm that can be considered as data-specific will be presented. I will show that both the dictionary and sparse codes can be effectively computed for several commonly encountered Riemannian manifolds. Experimental results are then presented for both reconstruction and classification problems from Computer Vision and Medical Image Analysis domains respectively. The preliminary results demonstrate that the dictionaries and the sparse codes learned using the proposed method can and do indeed provide real improvements when compared with other direct approaches.

Bio: Baba Vemuri received his PhD in Electrical and Computer Engineering from the University of Texas at Austin in 1987. He then joined the Department of Computer and Information Sciences at the University of Florida, Gainesville, where he is currently a full professor. His research interests include Medical Image Computing, Com-puter Vision, Machine Learning and Information Geometry. He has published over 180 refereed journal articles and conference proceedings in the aforementioned areas and won several best paper awards. He has served as a program chair and area chair of several IEEE conferences. He was an associate editor for several area journals and is cur-rently an associate editor for the Journal of Medical Image Analysis and the International Journal of Computer Vision (IJCV). Professor Vemuri is a fellow of the IEEE and ACM.

For more information, contact Prof. Guido Gerig.