Towards Practical Face Recognition System Employing Row-Based Distance Method In 2dpca Based Algorithms
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,...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/46292/1/Waled%20Hussein%20Mohammed%20Al-Arashi24.pdf http://eprints.usm.my/46292/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | 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. |
|---|