An Improved U-Net Architecture for Image Dehazing

In this paper, we present a simple yet powerful deep neural network for natural image dehazing. The proposed method is designed based on U-Net architecture and made some design changes to make it better. We first use Group Normalization replace Batch solve the problem of insufficient batch size due hardware limitations. Second, introduce FReLU activation into block, which can achieve capturing complicated visual layouts with regular convolutions. Experimental results public benchmarks demonstrate effectiveness modified components. On SOTS Indoor Outdoor datasets, obtains PSNR 32.23 31.64 respectively, are comparable performances state-of-the-art methods. code publicly available online soon.