Rong Chen

With the fast growth of deep learning, trainable frameworks have been presented to restore hazy images. However, the capability of most existing learning-based methods is limited since the parameters learned in an end-