Abstract: We propose a new novel algorithm to be able to implement fast and accurate search for the characteristic points in a face without constraint background in mobile environments. We modify the approach of Skin Color Model, which uses Hough Transformation to extract the fitting ellipse for detection of the facial feature, spent 90% of a whole running time. To improve the performance of detection, especially processing time, a simple geometric mathematics is applied. Then, Eigenface and Support Vector Machine (SVM) are performed on the subspace of detected face region to verify the identification of users. Applying simple geometric algorithm in extracting the facial feature and the Region of Interest (ROI) to Principal Component Analysis (PCA) are of benefit to much computation time are saved, cause to reduce the quantity of processed dataset. In experiment results, proposed algorithm shows that it will be more suitable for application in undefined environments, where varying lighting conditions with complex backgrounds, according to be fast and accurate than any other previous methods.
Yong-Hwan Lee , Tae-Kyu Han , Young-Seop Kim and Sang-Burm Rhee , 2005. An Efficient Algorithm for Face Recognition in Mobile Environments . Asian Journal of Information Technology, 4: 796-802.