Robust face recognition based on dynamic rank representation

• Dynamic rank estimation can extract optimal subspace which can recover the face images from corruption.
• Improves the efficiency of discriminative feature selection.
• Needs lower dimensions training samples but gains a higher recognition rate.
• Achieves competitive results with accuracy, robustness and efficiency.

Robust face recognition is an active topic in computer vision, while face occlusion is one of the most challenging problems for robust face recognition algorithm. The latest research on low-rank representation demonstrated its high efficiency to subspace segmentation and feature extraction. Motivated by previous work, in this paper, we consider the problem of human face recognition from frontal views with varying illumination, as well as occlusion and disguise. We present a novel approach for face recognition by extracting dynamic subspace of images and obtaining the discriminative parts in each individual. We use these parts to represent the characteristic of discriminative components, give a recognition protocol to classify face images. The experiments carried on publicly available databases (i.e., AR, Extended Yale B, and ORL) vilidate its accuray, robustness and speed. The proposed method needs lower dimensions training samples but gains a higher recognition rate than other popular approaches.