Single Image Dehazing Via A Joint Deep Modeling

It is of great research significance and application background to restore the potential clear image under a foggy image.On the one hand,there are many problems in computer vision,eg.target detection,pedestrian recognition,etc.,which regard the to be processed images or video frames as clear.However,due to the presence of fog in the atmosphere,the image taken is not very ideal,which brings great difficulties to the processing of some visual problems.On the other hand,image dehazing has received extensive attention from researchers in vision society.It is applied to many fields,such as military,traffic,art,etc.Previous dehazing methods usually estimate transmissions and haze-free images in a separate way,i.e.estimating the transmission and airlight firstly,then getting the haze-free image using Atmospheric Light Scattering Model,which leads to poor image dehazing results if transmissions are incorrectly estimated.On the other side,though some CNN-based deep networks have been developed to remove haze,their transmission estimations heavily rely on white balance.Contrastly,in this paper,we propose a residual type CNN for transmission refinement rather than estimation.Benefit from its residual learning ability,we plug the network in solving an optimization problem,which is able to improve the refinement results through jointly estimating transmissions and clean images in a single framework.This paper summarizes some mainstream image dehazing methods of recent years,including image dehazing method based on dark channel prior,deep learning and optimization,and analyzes their relative merits.On this basis,the relevant knowledge of deep learning and joint optimization is introduced.Eventually our joint deep modeling is demonstrated.Experimental results of synthetic and real-world images demonstrate the superiority and efficiency of our proposed framework,compared to many state-of-the-art methods.