Sorted locally confined non-negative matrix factorization in face verification

In this paper, we propose a face recognition technique based on modification of the nonnegative matrix factorization (NMF) technique, which is known as sorted locally confined NMF (SLC-NMF). The SLC-NMF uses NMF to find nonnegative basis images, a subset of which were selected according to a discriminant factor and then processed through a series of image processing operations to yield a set of ideal locally confined salient feature basis images. SLC-NMF illustrates a perfectly local salient feature region which effectively realizes the "recognition by parts" paradigm for face recognition. The best performance is attained by SLC-NMF compared to the PCA, NMF and local NMF, in the FERET face database.