Research And Improvement Of Image Haze Removal Algorithm Based On Dark Channel Priors And Its Application In Video Haze Removal

In many regions,due to environmental factors,images acquired by video devices contain a large number of foggy images.However,foggy images are blurry and have poor visual effects,and cover up a lot of key information.Therefore,it has important research values that make the fog images to be fog free images and increase the color saturation of the images and make the images contain more critical information.This thesis discusses the haze algorithm research and its application based on dark channel prior.And for foggy image,the algorithm is improved and applied to video image de-fogging.First of all,this thesis analyzes the major de-fogging technologies from the perspective of enhanced de-fogging and physical model de-fogging,and analyzes the dark channel prior to haze.Then,it is found that the algorithm has three main shortcomings:(1).The time and space complexity of the algorithm is high and the de-fogging efficiency is low;(2).After using the algorithm to haze,the sky area will be distorted,which will affect the visual effect;(3).The image after using the algorithm to haze has low color saturation and clarity,which is not conducive to extracting key information.Then,in view of the shortcomings of the algorithm,this thesis makes a targeted improvement:(1).Using Guide filtering and Bicubic interpolation to improve transmittance.the detailed transmittance is obtained,the time of acquiring transmittance is shortened,the efficiency of the algorithm is improved,and the image details are enhanced.(2).By referring to the Logarithmic function,the transmittance of the sky region is corrected,which increases the robustness of the algorithm to haze of the dark channel prior.(3).The methods that Histogram equalization was adopted to increase color brightness and Gaussian high-pass filtering to enhance details improve the color saturation and clarity of the image after de-fogging.Finally,the improved de-fogging algorithm is applied to video image de-fogging.In this thesis,the influence of variable parameters in the process of de-fogging is discussed,so as to solve the problem of how to select parameters.The similarity between video frames was also used to share the transmittance and atmospheric light value to improve the calculation efficiency of video defogging,and the experimental results verify the validity of the improved algorithm.