Research On Single Image Demogging Algorithm Based On Enhancement And Restoration

In foggy weather,the suspended particles in the atmosphere will scatter the outdoor images,resulting in the image quality reduction and the loss of details,which has a great impact on our work and life.Therefore,the fog removal of images or videos has become a research hotspot in computer vision.With the development of computer vision,image defogging has made a great breakthrough,but existing shortcomings of image defogging algorithm are color distortion,low contrast,detail loss,poor implementation and halo phenomenon in the sky.At present,the methods used in image defogging mainly include three kinds of methods: image enhancement,image restoration,and deep learning.Based on the study and exploration of the first two kinds of fog removal algorithms,improved methods are proposed for the shortcomings of the algorithms.The tasks of this thesis are as follows:(1)An improved Retinex algorithm based on sky segmentation is proposed.Firstly the image of the algorithm is stretched by contrast and counted RGB number of pixels,then the extreme values on the right side of cubic B-spline are used to determine the sky region segmentation prefabrication,finally,the color correction of the segmented non-sky region is done and the mean value of the sky region is processed.This method solves the halo problem in the sky area and the color deviation problem in the non-sky area.(2)An adaptive image defogging algorithm based on histogram equalization is proposed.Firstly,the image is converted into HSI color space,and the image is divided into k × k sub images.Then,on the basis of ensuring the H component of hue unchanged,the saturation S component and brightness I component are processed by adaptive histogram equalization.This method solves the problem of serious loss of edge details and low contrast.(3)An adaptive image defogging algorithm based on iteration is proposed.The algorithm adds constant C compensation to the algorithm based on dark channel prior theory.In order to protect the edge information,the bilateral filtering algorithm is adopted.Finally,the atmospheric light and transmittance are taken as a whole,and the iterative algorithm of optimization conditions is constructed combined with the peak signal-to-noise ratio to realize the adaptive image defoggingBoth subjective visual and objective quality evaluation parameters show the feasibility of the above three improved algorithms.