Data Augmentation for Face Recognition System Implemented in Multiple Transform Domains

Abstract

A face recognition system which represents each of the augmented facial images as a superposition of the dominant components in two transform domains is proposed. Each face in the spatial domain is divided into horizontal, vertical halves and diagonal format. These partitions are concatenated to generate four more faces per subject in any database used. All images are first preprocessed then compressed using two different domains. The Discrete Wavelet Transform (DWT) and the Discrete Cosine Transform (DCT). Accordingly, each face will have two feature matrices. A voting scheme is used to define ground truth identity. The performance of the proposed system is evaluated using k-fold cross validation of ORL, Yale and FERET databases. Sample results are presented. The proposed technique achieves higher recognition rates while retaining 74% savings in storage recently reported.