Jie Liang, University of Central Florida


Face Recognition using Compressive Sensing Ideas

Abstract: We propose a face recognition paradigm using reweighted L2 minimization with hashing, whose recognition rates are comparable to the random projection using L1 minimization. Yet, our method is not only much faster than the standard compressive sensing method of Yang et al [2], but also robust to occlusion. We show that the sparse solution can be recovered with a high probability because hashing preserves the restrictive isometry property [6] and reweighted L2 mirrors the L1 (i.e. L0) solution [1]. Moreover, we present a theoretical analysis on the convergence of the proposed L2 approach. Experiments show a very promising recognition rate even with occlusion and significant speedup compared with [2].

References:
[1] E. Candes, J. Romberg, and T. Tao. Stable Signal Recovery from Incomplete and Inaccurate Measurements. Communications on Pure and Applied Mathematics, p1207 -1233, 2005.
[2] A. Yang, J. Wright, Y. Ma, and S. Sastry. Feature Selection in Face Recognition: A Sparse Representation Perspective. Preprint, 2007.
[3] E. Candes, M. Wakin, and S.Boyd. Enhancing Sparsity by Reweighted L1 Minimization. The Journal of Fourier Analysis and Applications, 2004.
[4] I. Daubechies, R. DeVore, M. Fornasier, and C. Gunturk. Iteratively Re-weighted Least Squares Minimization for Sparse Recovery. Communications on Pure and Applied Mathematics, 2008.
[5] John Wright, Allen Yang, and Arvind Ganesh. Robust Face Recognition via Sparse Representation. IEEE Trans, 2008.
[6] Qinfeng Shi, Hanxi Li, and Chunhua Shen. Rapid Face Recognition Using Hashing. CVPR, 2010.

Mentors: Dr. Xin Li, Dr. Neils Lobo, and Dr. Mubarak Shah (Mathematics and Computer Science, UCF); Enrique Ortiz (Computer Vision Lab, UCF)