Video dehazing based on CNN

Abstract

The appearance of outdoor images is easily affected by natural phenomena such as fog and dust, which reduces contrast and color distortion. Video dehazing has a wide range of real-time applications, but the challenges mainly come from large amount of computation and bad real-time performance. In this paper, we propose a video dehazing system which is an end-to-end network based on CNN (Convolutional Neural Network). The dehazing algorithm learns the scene transmission and the global atmospheric light simultaneously, which simplifies the dehaze process and improves the real-time performance. Finally, we process videos through combining the end-to-end dehaze network and bicubic interpolation algorithm, and obtain satisfactory results. The experiment results demonstrate that the proposed method performs favorably against the state-of-the-art methods on both quantitative and qualitative evaluation.