Illumination-Robust Face Recognition from a Single Image per Person Using Matrix Polar Decomposition

Mehdi EZOJI  Karim FAEZ  

Publication
IEICE TRANSACTIONS on Information and Systems   Vol.E92-D   No.8   pp.1561-1569
Publication Date: 2009/08/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E92.D.1561
Print ISSN: 0916-8532
Type of Manuscript: PAPER
Category: Image Recognition, Computer Vision
Keyword: 
face recognition,  illumination invariant,  matrix polar decomposition,  wavelet domain,  single image per person,  

Full Text: PDF>>
Buy this Article




Summary: 
In this paper, a novel illumination invariant face recognition algorithm is proposed for face recognition. This algorithm is composed of two phases. In the first phase, we reduce the effect of illumination changes using a nonlinear mapping of image intensities. Then, we modify the distribution of the coefficients of wavelet transform in certain sub-bands. In this step, the recognition performance is more important than image quality. In the second phase, we used the unitary factor of polar decomposition of enhanced image as a feature vector. In the recognition phase, the correlation-based nearest neighbor rule is applied for the matching. We have performed some experiments on several databases and have evaluated the proposed method in different aspects. Experimental results in recognition show that this approach provides a suitable representation for overcoming illumination effects.