Enhancement of Low-Light Image Based on Wavelet U-Net

Abstract

In computer vision, low-light image enhancement has always been a challenging task caused by more lower signal to noise ratio. Some methods have been proposed to enhance the low-light image using fully convolution network. Using u-net as backbone, we introduce wavelet transform to conduct down-sampling and up-sampling operations. In order to recover more details, perceptual loss has been used to optimize the network parameters. Experiments show that our model can get better performance than the existing methods. We find that wavelet transform effectively improve the quality of low-light image enhancement.