Research On Single Image Dehazing Based On Atmospheric Scattering Model

With the rapid development of urbanization in China,more and more people are affected by haze.In the haze weather,the suspended solids in the atmosphere reflect,scatter and absorb light,which makes the image quality of image acquisition equipment seriously decline,leading to the lack of details and the decline of contrast of the image,which seriously affects the public transportation safety and outdoor video monitoring.Therefore,image dehazing technology becomes particularly important.In this paper,the problems of atmospheric light value estimation inaccuracy,sky color supersaturation and edge halo effect in the process of image dehazing are studied:(1)Aiming at the problem of the atmospheric light value estimation failure in the bright region and the image color supersaturation in the sky region in the dark channel prior,an image dehazing algorithm based on haze line prior is proposed.Firstly,the haze line prior and Hough transform are used to estimate the atmospheric light value,then the discrete Haar wavelet transform is used to construct the reduced dimension subband haze model,and the convex optimization regularization is used to refine the coarse transmission map.(2)Aiming at the problem of image color deviation and halo in the abrupt change of depth of field in the sky region due to the failure of transmission in the prior dark channel,an image dehazing algorithm is proposed which combines brightness model and gradient domain guided filtering.Firstly,the average value of the first 0.1% pixel with the largest brightness in the whole image is selected as the atmospheric light value;then,the improved dark channel model and brightness model are used to solve the transmission map of foreground area and sky region respectively,and the weighted fusion is used.Then,the gradient domain guided filtering of the first-order edge sensing constraint is used to refine it.(3)Aiming at the problems of blocking effect and halo effect caused by dark channel prior in depth discontinuity,an image dehazing algorithm based on weighted local constant assumption is proposed.The weight function is introduced to optimize the prior coarse transmission of the dark channel,so that it has more detailed edges;then the gradient domain guided filter of the first-order edge sensing constraint is used to refine.In order to verify the dehazing performance of the algorithm,Matlab software is used for simulation experiment,and the subjective evaluation and objective evaluation are compared with other algorithms.The experimental results show that the image dehazing effect of this algorithm is better,which can effectively solve the above problems,and the restored image without haze has higher definition,which also shows that this algorithm has more robustness and effectiveness.