This paper devises a new means of filter diversification, dubbed multi-foldfilter convolution (M-FFC), for face recognition. On the assumption that M-FFCreceives single-scale Gabor filters of varying orientations as input, thesefilters are self-cross convolved by M-fold to instantiate a filter offspringset. The M-FFC flexibility also permits cross convolution amongst Gabor filtersand other filter banks of profoundly dissimilar traits, e.g., principalcomponent analysis (PCA) filters, and independent component analysis (ICA)filters. The 2-FFC of Gabor, PCA and ICA filters thus yields three offspringsets: (1) Gabor filters solely, (2) Gabor-PCA filters, and (3) Gabor-ICAfilters, to render the learning-free and the learning-based 2-FFC descriptors.To facilitate a sensible Gabor filter selection for M-FFC, the 40 multi-scale,multi-orientation Gabor filters are condensed into 8 elementary filters. Asidefrom that, an average histogram pooling operator is employed to leverage the2-FFC histogram features, prior to the final whitening PCA compression. Theempirical results substantiate that the 2-FFC descriptors prevail over, or onpar with, other face descriptors on both identification and verification tasks.
展开▼