Multi-level dynamic error coding for face recognition with a contaminated single sample per person

Xiao Luan, Xin Wang, Linghui Liu, Weisheng Li

Abstract: Highlights•A new method of multi-level dynamic error coding is proposed for face recognition under single sample per person (SSPP).•A multi-level sparse representation scheme is used to exploit local and global information of reconstructed face errors.•A dynamic error coding method is designed to ignore the negative effects caused by the occlusion patch.•Our method shows strong generalization ability and is more robust against facial variations under SSPP.