Single Image Dehazing Algorithm Based On Weighted Guided Filtering And Adaptive Tolerance

In recent years,with the development of industrial and urbanization,the weather of haze has frequently appeared.The image acquired by the acquisition device under such harsh conditions may have problems such as reduced definition,low signal-tonoise ratio,and loss of key information of the image,which seriously affect the performances of the outdoor vision systems which include outdoor video surveillance systems,military fields systems,remote sensing monitoring systems,etc.Therefore,it is very important to achieve image defogging by practical methods.This paper first introduces the influence of atmospheric scattering on the imaging of objects,and uses the mathematical model to visually describe the cause of the degradation of the image.Then,the image defogging algorithm based on dark channel prior is introduced in detail.Dark channel priors are combined with atmospheric scattering models to calculate image transmission and atmospheric light.In view of the problem that the gross transmittance is estimated to cause a blockiness in the restored image,a guided filtering algorithm is used to refine the transmittance.At the same time,the value of atmospheric light is estimated by the quarter method in order to effectively reduce the error.Although the algorithm has good defogging ability,it still has some defects.The research work in this paper mainly includes the following contents:(1)Although that the guided filtering is used to refine the transmittance does not only reduce the time-complexity of the algorithm but also effectively deals with the block-ness caused by the coarse transmittance,the guided filtering algorithm uses a uniform linear model and regularization for each region of the entire image,which cannot adapt to the texture characteristics of different regions of the image itself,and may lead to halo artifacts in some edge regions of the restored image.For these problems,this paper designs an edge-aware weight function based on LOB operator in order to assign a larger weight to the pixels at the edge.The function is incorporated into the guided filter to form the weighted guided filter.Refining the transmittance with improved weighted guidance filters can effectively avoid halo artifacts at the edges.(2)When the foggy image has large areas such as sky,water surface and white scene,the recovery result of the algorithm may cause distortion such as color spots and color blocks.From the perspective of the atmospheric scattering model and the dark channel prior,it is found that the main cause of the distortion phenomenon is that the transmittance estimation of the bright region of the image is lower than the actual value.To solve this problem,this paper proposes a calculation method based on the characteristics of the foggy image itself to obtain the adaptive tolerance,which is used to correct the transmittance of the dark channel prior failure zone.The simulation results show that the improved algorithm can not only recover images with large areas of bright areas,but also avoid color distortion and better preserve image details.