Feature Selection in Face Recognition:

A Sparse Representation Perspective

Allen Y. Yang, John Wright,  Yi Ma, and Shankar Sastry

© Copyright Notice: It is important that you read and understand the copyright of the following software packages as specified in the individual items. The copyright varies with each package due to its contributor(s). The packages should NOT be used for any commercial purposes without direct consent of their author(s). 


We examine the role of feature selection in face recognition from the perspective of sparse representation. We cast the recognition problem as finding a sparse representation of the test image features w.r.t. the training set. The sparse representation can be accurately and efficiently computed by L-1 minimization. We show that if sparsity in the recognition problem is properly harnessed, the choice of features is no longer critical. What is critical is whether the number of features is sufficient and whether the sparse representation is correctly found. Our thorough evaluation shows that the proposed algorithm achieves much higher recognition accuracy on face images with variation in either illumination or expression. Furthermore, other unconventional features such as down-sampled images and randomly projected features perform almost equally well with the increase of the feature dimensions.

  John Wright, Allen Y. Yang, Arvind Ganesh, Shankar Sastry, and Yi Ma. Robust face recognition via sparse representation. To appear in PAMI, 2008. [Main, Supplemental]

Allen Y. Yang, John Wright, Yi Ma, and Shankar Sastry. Feature selection in face recognition: A sparse representation perspective. UC Berkeley Tech Report UCB/EECS-2007-99, 2007. [PDF]

MATLAB Toolbox: "Holistic Feature Selection for Face Recognition via Randomface Ensembles."

(c) Copyright. University of California, Berkeley. 2007.
Author: Allen Y. Yang. [yang AT eecs . berkeley . edu]

This technique is patent pending filed through UIUC and UC Berkeley IP offices. For licensing, please contact:
[Office of Intellectual Property & Industry Research Alliances]

Other Related L-1 Minimization Packages

Although we cannot freely release our source codes. Our algorithms are easy to implement with the following toolboxes.
Should you have any questions, we will also be glad to answer.
  • L-1 Magic. A MATLAB package developed at Caltech. [website]
  • SparseLab. Another MATLAB toolbox developed at Stanford. [website]


This technique has been cited in the following articles:
  • Rice University Compressive Sensing Resources. [link]
  • Monday Morning Algorithm Blog: Compressed Sensing Meets Machine Learning. [link]
  • Wired.com: Engineers Test Highly Accurate Face Recognition. [link]

[Return to Home]