Relative Depth Estimation Prior for Single Image Dehazing
Jinbao Wang, Ke Lü, Jian Xue, Yutong Kou · 2019
In this paper, a novel single image dehazing technique is proposed using relative depth estimation prior and based on an atmospheric scattering physical model. The contribution of our work is mainly that we put forward a prior named relative depth estimation prior to calculate the accurate value of transmission map, and a new iterative strategy has been proposed to predict the best dehazed result automatically in the hazy-free recovery process. In the qualitative and quantitative comparison part, we have compared our results with the state-of-the-art works, including traditional physical model based and convolutional neural network based methods and also test our method on the bechmark dataset. The experiments show our method is effective, and can enhence the image texture in details simultaneously.