Integrate the original face image and its mirror image
for face recognition
Yong Xu
a,b
, Xuelong Li
c,
n
, Jian Yang
d
, David Zhang
e
a
Bio-Computing Research Center, Shenzhen Graduate School, Harbin Institute of Technology, Shenzhen 518055, Guangdong, P.R. China
b
Key Laboratory of Network Oriented Intelligent Computation, Shenzhen 518055, Guangdong, P.R. China
c
Center for OPTical IMagery Analysis and Learning (OPTIMAL), State Key Laboratory of Transient Optics and Photonics, Xi'an Institute of Optics and Precision
Mechanics, Chinese Academy of Sciences, Xi'an 710119, Shaanxi, P.R. China
d
College of Computer Science and Technology, Nanjing University of Science and Technology, Nanjing, 210094, China
e
Biometrics Research Center, The Hong Kong Polytechnic University, Hong Kong
article info
Article history:
Received 3 June 2013
Received in revised form
21 September 2013
Accepted 22 October 2013
Communicated by D. Zhang
Keywords:
Pattern recognition
Biometrics
Face recognition
Mirror image
Sparse representation
abstract
The face almost always has an axis-symmetrical structure. However, as the face usually does not have an
absolutely frontal pose when it is imaged, the majority of face images are not symmetrical images. These
facts inspire us that the mirror image of the face image might be viewed as a representation of the face
with a possible pose opposite to that of the original face image. In this paper we propose a scheme to
produce the mirror image of the face and integrate the original face image and its mirror image for
representation-based face recognition. This scheme is simple and computationally efficient. Almost all
the representation-based classification methods can be improved by this scheme. The underlying
rationales of the scheme are as follows: first, the use of the mirror image can somewhat overcome the
misalignment problem of the face image in face recognition. Second, it is able to somewhat eliminate the
side-effect of the variation of the pose and illumination of the original face image. The experiments show
that the proposed scheme can greatly improve the accuracy of the representation-based classification
methods. The proposed scheme might be also helpful for improving other face recognition methods.
& 2013 Elsevier B.V. All rights reserved.
1. Introduction
As we know, the main challenges of face recognition are that
the face image might severely vary with the various poses, facial
expression and illumination [1–3]. A face recognition method
greatly suffers from these challenges. In order to address these
challenges, people have made many efforts. For example, Jian et al.
proposed the illumination compensation method for face recogni-
tion [4]. Sharma et al. proposed pose invariant virtual classifiers
for face recognition [5]. We also note that if the available training
samples of a face can sufficiently show possible variations of the
pose, facial expression and illumination, it will be possible to
obtain a high accuracy. Unfortunately, in real-world applications
a face usually has only a very small number of training samples,
which cannot convey many variations of the face [6–10]. In order
to overcome the problem that the training samples of a face do not
convey sufficient variations of the face, previous literatures
have proposed some approaches to generate new (i.e. virtual or
synthesized) face images and to enlarge the size of the set of the
training samples.
It is known that both the facial structure and the facial expr ession
are symmetrical [11]. Previo us literatures have successfully e xp loi te d
the symmetrical structure of the face for face detection [11–14].
Howev er, it should be pointed out that in real-w or ld face recogniti on
applications, a large number of face images are not symmetrical
images due to non-frontal and non-neutral pose [15].Xuetal.
proposed an approach to generate “symmetrical” face images and
exploited both the original and “symmetrical” face images to
recognize the subject [15].Asthe“symmetrical” face image is
gener ate d with the assumption that the facial structure is symme-
trical, it is an axis-symmetrical image. Howev er, as shown lat er, the
“symmetrical” face images obtained in [15] is not a natural face
image and even appear to be strange.
A real-world face recognition application also often suffers
from the misalignment problem of the face image. This problem
certainly makes the face image not symmetrical and is advanta-
geous for correctly recognizing the face. Fig. 3 presented later
shows an example of misalignment of the face image.
It should be pointed out that though previous literatures have
made many efforts in making virtual or synthesized face images
which reflect the variation of the face as much as possible, almost
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/neucom
Neurocomputing
0925-2312/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.neucom.2013.10.025
n
Corresponding author.
E-mail addresses: yongxu@ymail.com (Y. Xu), xuelong_li@opt.ac.cn (X. Li),
csjianyang@gmail.com (J. Yang), csdzhang@comp.polyu.edu.hk (D. Zhang).
Please cite this article as: Y. Xu, et al., Integrate the original face image and its mirror image for face recognition, Neurocomputing
(2013), http://dx.doi.org/10.1016/j.neucom.2013.10.025i
Neurocomputing ∎ (∎∎∎∎) ∎∎∎–∎∎∎