Learning sparse discriminant low-rank features for low-resolution face recognition

Abstract

In this paper, we propose a novel approach for low-resolution face recognition, under uncontrolled settings. Our approach first decomposes a multiple of extracted local features into a set of representative basis (low-rank matrix) and sparse error matrix, and then learns a projection matrix based on our proposed sparse-coding-based algorithm, which preserves the sparse structure of the learned low-rank features, in a low-dimensional feature subspace. Then, a coefficient vector, based on linear regression, is computed to determine the similarity between the projected gallery and query image's features. Furthermore, a new morphological pre-processing approach is proposed to improve the visual quality of images. Our experiments were conducted on five available face-recognition datasets, which contain images with variations in pose, facial expressions and illumination conditions. Experiment results show that our method outperforms other state-of–the-art low-resolution face recognition methods in terms of recognition accuracy.