A novel scheme for face recognition and authentication against pose, illumination and expression changes
Ying Nong Chen, Chin Chuan Han, Cheng Tzu Wang, Kuo Chin Fan · 2011
In this paper, a novel scheme for face recognition or authentication is proposed against pose, illumination, and expression (PIE) variation using modular face features. A sub-image in low-frequency sub-band is extracted by a wavelet transform (WT) to reduce the image dimensionality. It is partitioned into four parts for representing the local fea-tures and reducing the PIE effects, and the small image in a coarse scale is generated via the WT without losing the global face features. Five modular feature spaces were con-structed. The most discriminative common vectors in each feature space were found, and a nearest feature space-based (NFS-based) distance was calculated for classification. Fi-nally, a weighted summation is performed to fuse the five distances. Experiments were conducted to show that the proposed scheme is superior to other methods in terms of recognition and authentication rates.