SLR: Semi-Coupled Locality Constrained Representation for Very Low Resolution Face Recognition and Super Resolution

Abstract

Although face recognition algorithms have been greatly successful recently, in real applications of very low-resolution (VLR) images, both super-resolution (SR) and recognition tasks are more challenging than those in high-resolution (HR) images. Given the rare discriminative information in VLR images, the one-to-many mapping relationship between HR and VLR images degrades the SR and recognition performances. In this paper, we propose a novel semi-coupled dictionary learning scheme to promote discriminative and representative abilities for face recognition and SR simultaneously by relaxing coupled dictionary learning. Specifically, we use semi-coupled locality-constrained representation to enhance the consistency between VLR and HR local manifold geometries, thereby overcoming the negative effects of one-to-many mapping. Given the learned task-oriented mapping function, we feed these discriminative features into a collaborative representation-based classifier to output their labels, and combine a locality-induced approach to hallucinate the HR images. Extensive experimental results demonstrate that the proposed approach outperforms a number of state-of-the-art face recognition and SR algorithms.