Multi-Fold Gabor, PCA and ICA Filter Convolution Descriptor for Face Recognition

—This paper devises a new means of filter diversification, dubbed multi-fold filter convolution (í µí³œ-FFC), for face recognition. On the assumption that í µí³œ-FFC receives single-scale Gabor filters of varying orientations as input, these filters are self-cross convolved by í µí³œ-fold to instantiate an offspring set. The í µí³œ-FFC flexibility also permits the self-cross convolution amongst Gabor filters and other filter banks of profoundly dissimilar traits, e.g., principal component analysis (PCA) filters, and independent component analysis (ICA) filters, in our case. A 2-FFC instance therefore yields three offspring sets from: (1) Gabor filters solely, (2) Gabor and PCA filters, and (3) Gabor and ICA filters, to render the learning-free and the learning-based 2-FFC descriptors. To facilitate a sensible Gabor filter selection for í µí³œ-FFC, the 40 multi-scale, multi-orientation Gabor filters is condensed into 8 elementary filters. In addition to that, an average pooling operator is used to leverage the í µí¿-FFC histogram features, prior to whitening PCA compression. The empirical results substantiate that the 2-FFC descriptors prevail over, or on par with, other face descriptors on both identification and verification tasks.