Performance analysis of various distance measures for PCA based Face Recognition

Face Recognition (FR) is one the most active and widely investigated techniques in computer vision. It is used in a variety of problems like image and film processing, human-computer interaction, criminal identification and so on. The scheme is based on an information theory approach that decomposes face images into a small set of characteristic feature images called `eigenfaces´, which are essentially the principal components of the initial training set of face images. We perform FR using the popular dimensionality reduction technique, the Principal Component Analysis (PCA) and conduct extensive investigations on the ORL database to compare the three prominent distance classifiers: Euclidean, Mahalanobis and Cosine to determine which measure is more effective by stringently comparing them on the basis of the training set.