Please use this identifier to cite or link to this item: https://doi.org/10.1109/CVPR.2012.6247828
Title: Practical low-rank matrix approximation under robust L 1-norm
Authors: Zheng, Y.
Liu, G.
Sugimoto, S.
Yan, S. 
Okutomi, M.
Issue Date: 2012
Citation: Zheng, Y.,Liu, G.,Sugimoto, S.,Yan, S.,Okutomi, M. (2012). Practical low-rank matrix approximation under robust L 1-norm. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition : 1410-1417. ScholarBank@NUS Repository. https://doi.org/10.1109/CVPR.2012.6247828
Abstract: A great variety of computer vision tasks, such as rigid/nonrigid structure from motion and photometric stereo, can be unified into the problem of approximating a low-rank data matrix in the presence of missing data and outliers. To improve robustness, the L 1-norm measurement has long been recommended. Unfortunately, existing methods usually fail to minimize the L 1-based nonconvex objective function sufficiently. In this work, we propose to add a convex trace-norm regularization term to improve convergence, without introducing too much heterogenous information. We also customize a scalable first-order optimization algorithm to solve the regularized formulation on the basis of the augmented Lagrange multiplier (ALM) method. Extensive experimental results verify that our regularized formulation is reasonable, and the solving algorithm is very efficient, insensitive to initialization and robust to high percentage of missing data and/or outliers 1. © 2012 IEEE.
Source Title: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
URI: http://scholarbank.nus.edu.sg/handle/10635/84107
ISBN: 9781467312264
ISSN: 10636919
DOI: 10.1109/CVPR.2012.6247828
Appears in Collections:Staff Publications

Show full item record
Files in This Item:
There are no files associated with this item.

SCOPUSTM   
Citations

77
checked on Oct 14, 2018

Page view(s)

33
checked on Sep 14, 2018

Google ScholarTM

Check

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.