Harnessing Sparse and Low-Dimensional Structures for Robust Clustering of Imagery Data
Rao, Shankar Ramamohan
This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/81149
Description
Title
Harnessing Sparse and Low-Dimensional Structures for Robust Clustering of Imagery Data
Author(s)
Rao, Shankar Ramamohan
Issue Date
2009
Doctoral Committee Chair(s)
Ma, Yi
Department of Study
Electrical and Computer Engineering
Discipline
Electrical and Computer Engineering
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Keyword(s)
Computer Science
Language
eng
Abstract
We propose a robust framework for clustering data. In practice, data obtained from real measurement devices can be incomplete, corrupted by gross errors, or not correspond to any assumed model. We show that, by properly harnessing the intrinsic low-dimensional structure of the data, these kinds of practical problems can be dealt with in a uniform fashion. In particular, we propose two robust segmentation algorithms: an algebraic method for data from multiple quadratic manifolds, and an information-theoretic approach for data from multiple linear subspaces. Our techniques draw from many diverse areas, including lossy data compression, sparse representation, algebraic geometry, and robust statistics. We verify the efficacy of our algorithms by applying them to the segmentation of tracked image features of objects in a dynamic scene under the affine and perspective camera models, and the segmentation of a natural image into regions with homogeneous texture. We benchmark the performance of our methods on many publicly available notion and imagery datasets. Our results are on par with state-of-the-art results, in many cases exceeding them. Finally, we explore potential extensions and improvements to our techniques as well as new applications.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.
