Towards Practical Face Recognition System Employing Row-Based Distance Method In 2dpca Based Algorithms
Loading...
Date
2014-02
Authors
Al-Arashi, Waled Hussein Mohammed
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Automatic face recognition has been a focus research topic in past few decades.
This is due to the advantages of face recognition and the potential need for high security
in commercial and law enforcement applications. However, due to nature of
the face, it is subjected to several variations. Thus, finding a good face recognition
system is still an active research field till today. Many approaches have been proposed
to overcome the face variations. In the midst of these techniques, subspace methods
are considered the most popular and powerful techniques. Among them, eigenface or
Principal Component Analysis (PCA) method is considered as one of the most successful
techniques in subspace methods. One of the most important extensions of
PCA is Two-dimensional PCA (2DPCA). However, 2DPCA-based features are matrices
rather than vectors as in PCA. Hence, different distance computation methods
have been proposed to calculate the distance between the test feature matrix and the
training feature matrices. All previous methods deal with the classification problem
mathematically without any consideration between feature matrices and the face images.
Besides, the system performance in practical applications relies on the number
of eigenvectors chosen. As a solution to the above mentioned issues, four new distance
methods have been proposed in this thesis, which are based on the rows of a feature
matrix of 2DPCA-based algorithms. Through experiments, using eight face databases,
their improvements compared to the previous distance methods are demonstrated.
Description
Keywords
Face Recognition System , Row-Based Distance Method