| Previous | [ 1] | [ 2] | [ 3] | [ 4] | [ 5] | [ 6] | [ 7] | [ 8] | [ 9] | [ 10] | [ 11] | [ 12] | [ 13] | [ 14] | [ 15] | [ 16] | [ 17] | [ 18] | [ 19] | [ 20] | [ 21] | [ 22] | [ 23] | [ 24] |
¡@
YING-NONG CHEN1, CHIN-CHUAN HAN2, CHENG-TZU WANG3 AND KUO-CHIN FAN1
1Department of Computer Science and Information Engineering
National Central University
Taoyuan, 320 Taiwan
2Department of Computer Science and Information Engineering
National United University
Miaoli, 360 Taiwan
3Department of Computer Science
National Taipei University of Education
Taipei, 106 Taiwan
E-mail: cchan@nuu.edu.tw
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 features
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 constructed.
The most discriminative common vectors in each feature space were found, and
a nearest feature space-based (NFS-based) distance was calculated for classification. Finally,
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.
Received January 19, 2009; revised April 21 & June 23, 2009; accepted August 5, 2009.
Communicated by Tyng-Luh Liu.
* The work was supported by National Science Council of Taiwan under grant No. NSC 95-2221-E-239-035.