FACE RECOGNITION USING PCA WAVELET DECOMPOSITION 
O. O. Khalifa *, M. Yusuf, S. Gagula 
* Electrical & Computer Department, Kulliyyah of Engineering, 
International Islamic University Malaysia 
Jalan Gombak, 53100, Kuala Lumpur, Malaysia 
Fax: +603 2056 4853, Email: khalifa@iiu.edu.my, mumtaz2020@yahoo.com, sadina g@hotmail.com 
  
  
KEYWORDS: Biometrics, Extraction, Identification. Recognition, Automation. 
ABSTRACT 
Face recognition plays an important role in biometrics base personal identification. The need for reliable recognition and identification 
of interacting users is obvious. The biometrics recognition technique acts as an efficient method and wide applications in the area of 
  
information retrieval, automatic banking, control of access to security areas and s 
based on Principal Component A nalysis (PCA) and wavelet decomposition. The proposed scheme exploits feature extraction capabilities 
of the Discrete Wavelet Transform Decomposition and invokes certain normalization techniques that increase 
in facial geometry and illumination. Traditionally, to represent the human face, 
method, wavelet transform is used to decompose an image into diffe 
Jor PCA representation. In comparison with the traditional use 
discriminatory power; further, the proposed method reduces the 
existing algorithms. 
1. INTRODUCTION 
Face recognition has become and active area of research as it 
plays an important role in biometrics base personal identification. 
Thus, the need for reliable recognition and identification of 
interacting users is obvious. Human face recognition finds its 
application in law enforcement and commercial applications. 
Some of these applications are: static matching of controlled 
format photographs such as passports, credit cards, photo ID's, 
driver's licenses, and mug shots, as well as dynamic matching 
(real time matching) of video images (Zhang, 2000). 
In spite of this, face recognition technology seems to be a difficult 
task to develop since the appearance of a face varies dramatically 
because of illumination, facial expression, head pose, and image 
quality determine the recognition rate. In addition, the number of 
the same face in the database with different facial expression 
should be sufficient so that the person can be recognized in all 
possible situations. The recent research on face recognition is 
based on Principal Component Analysis (PCA). However, any 
system in this world has its limitations and can be improved. To 
overcome the disadvantages of PCA, such as large computational 
load and low discriminatory power it can be combined with 
Wavelet Transform (Chen ef al, 2003). 
A system capable of recognizing faces with different orientations 
and facial expression base on PCA and Wavelet decomposition 
was developed in fully using MATLAB. A review of basic 
fundamental of Principal Component Analysis and Wavelet 
Decomposition are introduced. Experimental results using 10 
images with five orientations are shown. The accuracy and 
performance of the system also presented. 
780 
2. PRINCIPLE COMPONENT ANALYSIS (PCA) 
PCA is a statistical measurement method, which operates in the 
linear domain and can be used to reduce the dimensionality of an 
image. A face image can be viewed as vectors and represented in 
matrix form. This method can be described as follows: 
Suppose A-[ai],. is a face image, where r and c are the number 
of rows and column of the images, respectively; aj is the grey 
value of the pixel in the i" row and j^ column. This matrix can be 
arranged into a column vector: 
X 7 [aii 201,81 812 82»... ap …. 1e re … ME 
where X is a D 7 rxc dimension vector. 
One face image can be considered as statistical sample. Thus, 
giving a group of face image samples in the training database, G — 
| Xo, X, ..., Xy, ) and the covariance matrix can be calculated as 
1 M-1 T 
Sz— Xi — my Xi — m : 
o X ) 
where m is the average vector of the training samples and M is the 
number of images in the training sample set. 
1 M -1 
Let Àj A» e A. and' as do, oa Ha be eigenvalues and 
corresponding eigenvector obtain from the covariance of S 
o no. This paper describes a method of face recognition 
its robustness to variations 
PCA is performed on the whole facial image. In this 
rent frequency subbands, and a mid-range frequency subband is used 
of PCA, the proposed method gives better recognition accuracy and 
computational load significantly when the image database is large. 
Experimental results show that the proposed method is effective and possesses several desirable properties when it compared with many 
m N'Y Pp