Image Dehazing Based On Variational Partial Differential Equations

The albedo effect of small particles such as impurities and particulate matter suspended in the air and water on atmospheric light can cause interference to outdoor and underwater image information acquisition systems,and the obtained outdoor and underwater images often have problems such as haze,noise,low contrast and poor clarity.This causes serious difficulties for subsequent image processing,such as segmentation and recognition.Therefore,in practical applications,there is an urgent need for effective image dehazing and denoising methods to restore clear haze-free images.This paper studies the image dehazing methods based on partial differential equation.The main research contents and innovations are as follows:(1)The existing dark channel prior and underwater dark channel prior are not accurate for estimating the transmission of the bright areas of the images,which results in the haze-free images distortion,darkening and edge detail blurring.In this paper,we introduce the dark channel prior with edge indicator and the underwater dark channel prior with edge indicator for estimating initial scene depth.The experimental results show that the two priors can well protect the edge details of the scene depth,which lays the foundation for the final restoration of clear haze-free images.(2)By combining the dark channel prior with edge indicator and the underwater dark channel prior with edge indicator to estimate the initial scene depths,we propose two variational dehazing models to restore outdoor images and underwater images respectively.Furthermore,the existence and uniqueness of the minimizers to the proposed two models are established.(3)By incorporating the primal-dual method,we design efficient algorithms for solving the proposed models numerically and give the convergence analysis of the algorithms.Experimental results show that the variational dehazing model and its algorithm based on outdoor images can quickly restore haze-free images with clear edges and high contrast.The variational dehazing model and its algorithm based on underwater images can well preserve the edge information of underwater images,correct the contrast of underwater images and obtain high-quality haze-free images.