Face Recognition Using Contourlet Transform and Multidirectional Illumination from a Computer Screen - Advanced Concepts for Intelligent Vision Systems - page 331

Face Recognition Using Contourlet Transform
and Multidirectional Illumination from a
Computer Screen

Ajmal Mian

School of Computer Science and Software Engineering
The University of Western Australia
35 Stirling Highway, Crawley, WA 6009, Australia
[e-mail]

Abstract. Images of a face under arbitrary distant point light source
illuminations can be used to construct its illumination cone or a linear
subspace that represents the set of facial images under all possible illu-
minations. However, such images are dicult to acquire in everyday life
due to limitations of space and light intensity. This paper presents an
algorithm for face recognition using multidirectional illumination gener-
ated by close and extended light sources, such as the computer screen.
The Contourlet coecients of training faces at multiple scales and ori-
entations are calculated and projected separately to PCA subspaces and
stacked to form feature vectors. These vectors are projected once again
to a linear subspace and used for classification. During testing, similar
features are calculated for a query face and matched with the training
data to find its identity. Experiments were performed using in house
data comprising 4347 images of 106 subjects and promising results were
achieved. The proposed algorithm was also tested on the extended Yale B
and CMU-PIE databases for comparison of results to existing techniques.

1 Introduction

Face recognition under varying illumination is a challenging problem because the
appearance of a face changes dramatically with illumination. In fact, changes
due to illumination can be greater than the changes due to face identity. Other
variations due to pose and facial expressions can introduce further challenges
however, they are less problematic when the subject is cooperative. In this paper,
we consider variations due to illumination alone.
Face recognition is extensively studied due to its potential applications in se-
curity, surveillance and human computer interaction. Zhao et. al. [26] provide a
detailed survey of face recognition literature and categorize them into holistic
face recognition techniques which match global features of the complete face
[23][3], feature-based techniques which match local features of the face [25] and
hybrid techniques which use both holistic and local features. From the perspec-
tive of data, face recognition can be divided into appearance based or shape
based techniques. While appearance based techniques are considered sensitive