Face Video Super-Resolution with Identity Guided Generative Adversarial Networks

Dingyi Li, Zengfu Wang

Abstract: Faces are of particular concerns in video surveillance systems. It is challenging to reconstruct clear faces from low-resolution (LR) videos. In this paper, we propose a new method for face video super-resolution (SR) based on identity guided generative adversarial networks (GANs). We establish a two-stage convolutional neural network (CNN) for face video SR, and employ identity guided GANs to recover high-resolution (HR) facial details. Extensive experiments validate the effectiveness of our proposed method from the following aspects: fidelity, visual quality and robustness to pose, expression and illuminance variations.