Modelling Uncertainty in Representation of Facial Features for Face Recognition

Abstract

Face is one of the important biometric identifier used for human recognition. The face recognition involves the computation of similarity between face images belonging to the determination of the identity of the face. The accurate recognition of face images is essential for the applications including credit card authentication, passport identification, internet security, criminal databases, biometric cryptosystems etc. Due to the increasing need for the surveillance and security related applications in access control, law enforcement, and information safety due to criminal activities, the research interest in the face recognition has grown considerably in the domain of the pattern recognition and image analysis. A number of approaches for face recognition have been proposed in the literature (Zhao et al. 2000), (Chellappa et al. 1995). Many researchers have addressed face recognition based on geometrical features and template matching (Brunelli and Poggio, 1993). There are several well known face recognition methods such as Eigenfaces (Turk and Pentland 1991), Fisherfaces (Belhumeur et al. 1997), (Kim and Kitter 2005), Laplacianfaces (He et al. 2005). The wavelet based Gabor function provide a favorable trade off between spatial resolution and frequency resolution (Gabor 1946). Gabor wavelets render superior representation for face recognition (Zhang, et al. 2005), (Shan, et al. 2004), (Olugbenga and Yang 2002). In recent survey, various potential problems and challenges in the face detection are explored (Yang, M.H., et al., 2002). Recent face detection methods based on data-driven learning techniques, such as the statistical modeling methods (Moghaddam and Pentland 1997), (Schneiderman, and Kanade, 2000), (Shih and Liu 2004), the statistical learning theory and SVM based methods (Mohan et al., 2001). Schneiderman and Kanade have developed the first algorithm that can reliably detect human faces with out-of-plane rotation and the first algorithm that can reliably detect passenger cars over a wide range of viewpoints (Schneiderman and Kanade 2000). The segmentation of potential face region in a digital image is a prelude to the face detection, since the search for the facial features is confined to the segmented face region. Several approaches have been used so far for the detection of face regions using skin color information. In (Wu, H.Q., et al., 1999), a face is detected using a fuzzy pattern matching method based on skin and hair color. This method has high detection rate, but it fails if the hair is not black and the face region is not elliptic. A face detection algorithm for color images using a skin-tone color model and facial features is