06-05-2011, 10:15 AM
Abstract
Together with the growing interest in the development of human and computer interface and
biometric identification, human face recognition has become an active research area since early
90. Nowadays, Principal Component Analysis (PCA) has been widely adopted as the most
promising face recognition algorithm. Yet still, PCA has its limitations: poor discriminatory power
and large computational load.
In view of these limitations, this paper proposed a new approach in using PCA - apply PCA on
wavelet subband. Traditionally, to represent the human face, PCA is performed on the whole facial
image. In the proposed method, wavelet transform is used to decompose an image into different
frequency subbands, and a mid-range frequency subband is used for PCA representation. In
comparison with the traditional use of PCA, the proposed method gives better recognition accuracy
and discriminatory power; further, the proposed method reduces the computational load
significantly when the image database is large, with more than 256 training images. This paper
details the design and implementation of the proposed method, and presents the encouraging
experimental results.
Keywords: Human face recognition, Principal Component Analysis, Wavelet transform, Subband.
1. Introduction
Over the past 20 years, extensive research works on various aspects of face recognition by human
and machines [1-3, 8-12, 16-28] have been conducted by psychophysicists, neuroscientists and
engineering scientists. Psychophysicists and neuroscientists have studied issues such as uniqueness
of faces, how infants perceive faces and organization of memory of faces. While engineering
scientists have designed and developed face recognition algorithms. This paper continues the work
done by engineering scientists in face recognition by machine.
Automatic face recognition by computer can be divided into two approaches [1,2], namely,
constituent-based and face-based. In constituent-based approach, recognition is based on the
relationship between human facial features such as eyes, mouth, nose, profile silhouettes and face
boundary [6,10, 27, 31]. The success of this approach relies highly on the accuracy of the facial
feature detection schemes. However, extracting facial features accurately is difficult. Every human
face has similar facial features, a small derivation in the extraction may introduce a large
classification error.
Face-based approach [13, 23, 25, 27, 28] attempts to capture and define the face as a whole. The
face is treated as a two-dimensional pattern of intensity variation. Under this approach, face is
matched through identifying its underlying statistical regularities. Principal Component Analysis
(PCA) [13, 16, 21] has been proven to be an effective face-based approach. Sirovich and Kirby [21]
first proposed using Karhunen-Loeve (KL) transform to represent human faces. In their method,
faces are represented by a linear combination of weighted eigenvector, known as eigenfaces. Turk
and Pentland [23] developed a face recognition system using PCA.
However, common PCA-based methods suffer from two limitations, namely, poor discriminatory
power and large computational load. It is well known that PCA gives a very good representation of
the faces. Given two images of the same person, the similarity measured under PCA representation
is very high. Yet, given two images of different persons, the similarity measured is still high. That
means PCA representation gets a poor discriminatory power. Swets and Weng [25] also observed
this drawback of PCA approach and further improve the discriminability of PCA by adding linear
discriminant analysis (LDA). But, to get a precise result, a large number of samples for each class
is required. On the other hand, O'Toole et al. [17] proposed different approach for selecting the
eigenfaces. They pointed out that the eigenvectors with large eigenvalues are not the best for
distinguishing face images. They also demonstrated that although the low dimensional
representation is not optimal for recognizing a human face, gives good results in identifying
physical categories of face, such as gender and race. However, O’Toole et al. have not addressed
much on the selection criteria of eigenvectors for recognition
Download full report.
http://citeseerx.ist.psu.edu/viewdoc/dow...1&type=pdf