Wavelet U-Net and the Chromatic Adaptation Transform for Single Image Dehazing

Abstract

Deep learning has played an important role in various computer vision tasks, and so does in single image dehazing. We think edges and colors are two key factors to obtain better dehazed images. Clear edges and balanced color make the dehazed images look natural and detailed. Therefore, we establish a novel two-stage and end-to-end network that comprises two characteristics. First, we propose the wavelet U-net, which uses discrete wavelet transform and inverse discrete wavelet transform to replace down-sampling and up-sampling, to extract edge features. Second, we apply the chromatic adaptation transform which can be implemented by convolutional layers mathematically to enhance images. Our neural network combining two concepts outperforms the state-of-art algorithms in visual performance and quantitative aspects.