Deep Dive: An End-to-End Dehazing Method Using Deep Learning
Image dehazing can be described as the problemof mapping from a hazy image to a haze-free image. Mostapproaches to this problem use physical models based onsimplifications and priors. In this work we demonstrate that aconvolutional neural network with a deep architecture and a largeimage database is able to learn the entire process of dehazing,without the need to adjust parameters, resulting in a much moregeneric method. We evaluate our approach applying it to realscenes corrupted by haze. The results show that even though ournetwork is trained with simulated indoor images, it is capable ofdehazing real outdoor scenes, learning to treat the degradationeffect itself, not to reconstruct the scene behind it.