The problem of face recognition is a composite task that involves the detection and location of faces in a cluttered background, normalization, recognition and verification. Depending on the nature of the application like the sizes of the training and testing databases, clutter and visibility of the background, noise, occlusion and speed requirements, some of these tasks could be very challenging. There have been several methods proposed for face recognition. And one of the key components of any methods is facial feature extraction. There are two major approaches to facial feature extraction for recognition, holistic template matching based systems and geometrical local feature based systems. In this paper we present the methods of PCA and LDA which are based on the holistic approach. The paper first focuses the major aspects of the implementation of both PCA and LDA and finally compare the performance analysis of both the techniques on the standard Image data base of AT&T