From Night to Day: GANs Based Low Quality Image Enhancement

Abstract

Nighttime images have lower contrast and higher noise than their daytime counterparts of the same scene. The aim of nighttime image enhancement is to improve the visual quality of nighttime images, so that they are visually as close as possible to daytime images. This problem is still challenging because of the deteriorated conditions of illumination lack and uneven lighting. In this paper, we propose a generative adversarial networks (GANs) based framework for nighttime image enhancement. To take advantage of GANs’ powerful ability of generating image from real data distribution, we make the established network well constrained by combining several loss functions including adversarial loss, perceptual loss, and total variation loss. Particularly, a pre-trained network is applied to leverage the perceptual loss which is beneficial to generate high-quality images. Meanwhile, for tackling the light-at-night effect, we present a fusion network in which the dark channel prior based illumination compensation is employed for the training of generator network. Experimental results have demonstrated the effectiveness of the proposed nighttime image enhancement network.