A Brief Review of Image Dehazing Algorithms Based on Deep Learning

Single image dehazing is a challenging problem, and it is far from solved. The application of deep learning in dehazing is only in the initial stage of exploration since the structure of deep learning is not designed for it. It occurs frequently that outdoor image quality is seriously affected when capturing image outside with dense haze, the contrast of the picture drops, and the information is lost due to the particles in the atmosphere. It seems indispensable to work on images without dehazing. A great number of methods have been proposed over the past dozen years. Those methods can be divided into traditional and deep learning methods. This paper mainly summarizes and uses traditional algorithms to compare, explains the classic algorithms of deep learning and introduces recent new efficient algorithms. The deep learning method architecture in the paper has been classified into the following two categories, (a) Convolution neural network (CNN) and (b) Generative adversarial network (GAN).