Abstract
This study performs reliability analysis on the different facial features with weighted retrieval accuracy on increasing facial database images. There are many methods analyzed in the existing papers with constant facial databases mentioned in the literature review. There were not much work carried out to study the performance in terms of reliability and also how the method will perform on increasing the size of the database. In this study certain feature extraction methods were analyzed on the regular performance measure and also the performance measures are modified to fit the real time requirements by giving weight ages for the closer matches. In this study four facial feature extraction methods are performed, they are DWT with PCA, LWT with PCA, HMM with SVD and Gabor wavelet with HMM. Reliability of these methods are analyzed and reported. Among all these methods Gabor wavelet with HMM gives more reliability than other three methods performed. Experiments are carried out to evaluate the proposed approach on the Olivetti Research Laboratory (ORL) face database.
Highlights
The main objective of face image retrieval is to get the ranking result from most to least similar face images in a face image database given a query face image
The algorithm in this study uses the low frequency sub images formed in the 2-D Discrete Wavelet Transform instead of the original face images, so that the dimension of the total population scatter matrix would reduced in the character extraction in the Principal Component Analysis (PCA) method and lessen the calculation amount
We performed our simulations on the Olivetti Research Laboratory (ORL) database with the platform of MATLAB 7.8
Summary
The main objective of face image retrieval is to get the ranking result from most to least similar face images in a face image database given a query face image. The algorithm in this study uses the low frequency sub images formed in the 2-D Discrete Wavelet Transform instead of the original face images, so that the dimension of the total population scatter matrix would reduced in the character extraction in the PCA method and lessen the calculation amount. We construct the M×M matrix L = ATA, where Lmn = Φ Φ and find the M eigenvectors, vi, of L These vectors determine linear combinations of the M training set face images to form the eigenfaces UI:. Without loss of generality N (O, μik, Uik) is assumed to be a Gaussian PDF with mean vector μik and covariance matrix Uik. Singular value decomposition: Singular value Decomposition methods for Face retrieval use the common result stated by the following theorem.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Research Journal of Applied Sciences, Engineering and Technology
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
