Haze Image Analysis And Simulating Application Research

In this paper,a large number of images were collected in different atmosphere conditions,and the atmospheric parameters are recorded synchronously.The relationship between the quality of haze images and the atmospheric parameters was studied.Image depth information is obtained using dark primary color theory,and no haze images were added haze.Aiming at the problem of residual and efficiency in the halo removal of the guided filtering algorithm in the process of dark channel prior,double-windows filtering algorithm was proposed,which improves the efficiency of the operation.The basic characteristics of fog and haze were analyzed.The scattering characteristic of single sphere particles was analyzed based on the Mie theory.The haze weather imaging mode applied to image simulation was described.Through the analysis of evaluation indexes of haze image quality,the edge strength of the image was chosen as the bisis for measuring the degree of the haze in image.By analyzing the relationship between the edge strength of image and the atmospheric parameters,it was found that AQI and the mass concentrations of PM_2.5 and PM₁₀ are closely related to the edge strength of the image,which is consistent with the exponential decay rule.The depth of field information of no haze image was expressed by dark channel prior in the form of transmittance,and it is the basis for the image haze simulation.The pavement area of the transmittance image was extracted by the maximum interclass variance method,and its transmittance was modified.Images were divided into five grades based on the mass concentration of PM_2.5.The transmittances of no haze images and various levels of haze images were refined and smoothed by guided filter.The empirical parameter of each grade was summarized by summing up the relationship between no haze image transmittance and the haze image transmittances at each level.Combining the change rule of the whole atmospheric light intensity of each grade images,no haze image was added haze.Halo effect occurs in areas where the depth of the scene changes greatly.The ability of the guided filtering algorithm often used in dark channel prior to remove halo is limited.In this paper,the cause of the halo effect was analyzed,and double-windows filtering algorithm based on dark channel prior was proposed to remove the halo effect.The halo region was identified by the difference between the large window transmittance and the small window transmittance,and its transmittance was corrected by linear superposition of the two transmittances.The superposition coefficient was adjusted by the difference in transmittance.Gaussian low-pass filter was used to smooth the corrected transmittance.Double-windows filtering algorithm effectively removes the halo effect and improves the computational efficiency compared to the guided filter algorithm.