Share Email Print
cover

Proceedings Paper

Wavelet decomposition-based efficient face liveness detection
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Existing face recognition systems are susceptible to spoofing attacks. So, Face liveness detection is a pivotal part for reliable face recognition, which has recently acknowledged vast attention. In this paper we propose a wavelet decomposition based face liveness recognition system using an energy calculation technique. Live faces contain high energy components compared to fake or printed image. In this paper, we calculate energy components of live face as well as fake face using discrete wavelet decomposition method. We analyze percentage of energy at different levels as well as for different wavelet basis function. We also analyze percentage of energy at different RGB bands and efficient face liveness detection method has been proposed. Discrete wavelet representation has been used to calculate decomposed energy components. Moreover, it provides differentiation of several spatial orientations as well as average and detailed information which are missing in the fake faces. This technique provides excellent discrimination capability when compared to the previously reported works based on the discrete Fourier transform and n-dimensional Fourier transform operations. To verify the proposed approach, we tested the performance using various face antispoofing datasets such as university of south Alabama (UFAD), and MSU face antispoofing dataset which incorporates different types of attacks. The test results obtained using the proposed technique shows better performance compared to existing techniques.

Paper Details

Date Published: 20 April 2016
PDF: 13 pages
Proc. SPIE 9845, Optical Pattern Recognition XXVII, 98450J (20 April 2016); doi: 10.1117/12.2224181
Show Author Affiliations
Md. Moniruzzaman, Univ. of South Alabama (United States)
Mohammad S. Alam, Univ. of South Alabama (United States)


Published in SPIE Proceedings Vol. 9845:
Optical Pattern Recognition XXVII
David Casasent; Mohammad S. Alam, Editor(s)

© SPIE. Terms of Use
Back to Top