The Research And Application Of Color Ellipsoid Prior Dehazing Algorithm Based On AnisGF

The problems such as blurring,low contrast and color distortion often exist in outdoor images obtained in bad environment such as fog and haze,which not only greatly limits the use of outdoor image acquisition system,but also impedes the subsequent analysis and application of images.Therefore,it is very important to recover high-quality clear images by image dehazing technology,which has become a hot research topic in civil and military fields,and has been widely concerned by scholars at home and abroad.In this paper,I first analyze the causes of hazy image quality degradation according to the atmospheric scattering model;then deeply study the single image dehazing algorithm based on color ellipsoid prior,and propose a new optimization algorithm for its problems;finally,a large number of dehazing experiments are carried out on outdoor hazy images and coal burning flame images of rotary kilns.The main research contents are as follows:(1)A method for calculating and refining the transmission value based on Anis GF is proposed to solve the problem of inaccuracy in calculating the transmission value of the single image dehazing algorithm based on color ellipsoid prior.As a weak anisotropic filter,the guided image filter is inaccurate in the calculation of transmission value due to the interference of noise when calculating the fuzzy segmentation weight.Therefore,the weighted average operation of the Anis GF with stronger anisotropy and better edge retention characteristics can be used to recalculate the more accurate transmission value,which can effectively suppress the halo phenomenon in the edge region of the dehazing images.(2)An adaptive quadtree method for estimating atmospheric light is proposed to solve the problem of inaccurate estimation of atmospheric light by the single image dehazing algorithm based on color ellipsoid prior.The median filtering operation is carried out on the hazy image to avoid the interference of highlighted objects in the image,the difference between the average and standard deviation of all pixel points in the four regions is used as the basis for the quadtree segmentation,and the adaptive threshold method during segmentation is used to prevent the estimation of atmospheric light value from being inaccurate,finally,the maximum intensity value in the selected area is used as the atmospheric light value.The atmospheric light value obtained by the adaptive quadtree estimation method is more accurate and faster than that obtained by the traditional estimation method.(3)Aiming at the problem that the coal burning flame images are blurred and unevenly illuminated by the flame area and dust particles in the rotary kiln,a defogging algorithm for the coal burning flame image based on local information is proposed.The algorithm can estimate the atmospheric light value at different positions according to the characteristics of the flame light source unique to the coal burning flame images,which solves the problem that the atmospheric light value of the coal burning flame image is not a global constant.The algorithm not only has a better enhancement effect,and effectively eliminates the interference of dust particles on the coal burning flame images,but also has a faster running speed,which can meet the requirements of real-time image dehazing in practical applications.