Classification of high-energized Gabor responses using Bayesian PCA for Human Face Recognition

Feature extraction methods are based on finding fiducial points (or local small areas) on a face and representing corresponding information in an efficient way. In this paper a novel method is proposed, based on selecting peaks (high-energized points) of the Gabor wavelet responses as feature points. Feature vectors are constructed from these feature points to represent the facial topography. These extracted feature vectors are used to obtain similarity scores of different features, using Bayesian Principal Component Analysis (PCA). Use of these peaks forms the basis of the work. These similarity scores are used for class assignment. Our approach not only reduces computational complexity, but also improves the performance in the presence of occlusions. As both Gabor filter responses and Bayesian Principal Component Analysis reduces intrapersonal variations, both the methods are integrated for better accuracy of face recognition. The efficiency of our method is demonstrated by the experiment on 1000 images from the FRAV2D face database where the images vary in pose, expression, illumination and scale, and 400 images from the ORL face database, where the images slightly vary in pose. Our method has shown 99% recognition accuracy for the FRAV2D database and 100% for the ORL database.