Research On Atmospheric Scattering Model And Point Light Source Multi-scattering Model For Image Restoration

With the rapid development of China's economy and society,the degree of environmental pollution has increased,leading to the frequent occurrence of smoggy weather throughout the country in recent years.The smog not only directly harms human health,but also reduces the image quality of various monitoring systems,which affects the smooth progress of related work.It can be seen that the restoration of this kind of image has high application value and economic value,and it is a research hotspot in the field of computer vision and imaging.For daytime smog images,the atmospheric scattering model has a color distortion and over-enhancement after the restoration of the sky region.A new method for estimating the atmospheric light is proposed,that is,according to the perspective principle,the farthest image is found at the intersection of the sky and the non-sky.The area is the heaviest foggy area,and the gray level mean of each channel of the RGB three channels in this area is used as an estimate of atmospheric light.After the defogging recovery,the phenomenon of color distortion or over-enhancement occurs in the foreground and sky regions,and the estimation of the transmittance is optimized.According to the change of the scene distance,the set factor is used to control the transmittance of' different scene distances,and the method based on context regularization is used instead of the guide filter to optimize the transmittance.Finally,the transmittance of the sky region is limited to prevent the sky color distortion after defogging..By testing many types of images and performing subjective and objective analysis and comparison with existing mainstream algorithms,the results show that the proposed algorithm has the advantages of natural color,low noise and high color recovery.For the night fog image,the image detail after the restoration of the atmospheric scattering model is not fine enough.And the light halo is not suppressed,even enlarged.This paper analyzes the principle of light imaging and finds that the model does not consider all factors of the image degradation.So a new point source multi-scatter recovery model is proposed.Specifically,each pixel in the nighttime image is regarded as a point source,and the atmospheric point spread function is used to simulate the multi-scattering of the point source to eliminate the influence on the imaging and achieve a better restoration effect.By testing many kinds of images and performing subjective and objective analysis and comparison with the existing mainstream algorithms,the results show that the proposed algorithm can better suppress street lights,lights and other lighting halos while ensuring the restoration of nighttime image details.Since it is not necessary to estimate atmospheric light and does not need to calculate accurate transmittance,it is 10 to 15 times faster than other restoration models and enhancement algorithms.