Face recognition using m-MSD and SVD with single training image

Maximum scatter difference (MSD) has been widely used in face recognition for feature extraction. However, its advantage will decrease when each object has only one training sample because the intra-class variations cannot be statistically measured in this case. To address the problem, a novel method based on m-MSD and SVD is proposed in this paper. A facial image is decomposed by the SVD algorithm, so one image can be transformed into several approximate images by reconstructing method with different number of singular values. That is to say, the number of training sample for each object is increased by singular value decomposition algorithm. Thus, the MSD algorithm can be applied to extract the discriminant features. Experiment results based on FERET and ORL face database show that the proposed method is efficient and it can achieve higher recognition rate than several existing algorithms.