Learning robust and discriminative low-rank representations for face recognition with occlusion

作者：

Highlights：

• We focus on face recognition scenarios where both training and testing image are corrupted duo to occlusions.

• We propose to learn robust and discriminative representation based on low-rank matrix recovery model.

• We solve the proposed model by using ALM, and the complexity analysis and convergence analysis are also provided.

• Experimental results demonstrate the effectiveness of our method.

摘要

•We focus on face recognition scenarios where both training and testing image are corrupted duo to occlusions.•We propose to learn robust and discriminative representation based on low-rank matrix recovery model.•We solve the proposed model by using ALM, and the complexity analysis and convergence analysis are also provided.•Experimental results demonstrate the effectiveness of our method.

论文关键词：Face recognition,Low-rank matrix recovery,Nuclear norm

论文评审过程：Received 8 July 2016, Revised 19 December 2016, Accepted 20 December 2016, Available online 21 December 2016, Version of Record 12 March 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.12.021