Study On Defogging Method Of Single Foggy Image

With the development of machine vision technology,image is more and more widely used in people's daily life.However,the images collected outdoors are easily affected by foggy weather,resulting in poor contrast,low definition,and color distortion of the acquired images,which greatly limits the effectiveness of the outdoor vision system.Therefore,the research on image defogging has important practical significance.Based on the analysis of human visual characteristics and foggy image degradation model,this thesis studies the single image defogging method from two directions: image enhancement and image restoration.In the field of image enhancement,the image defogging algorithm based on Retinex theory is mainly studied.The influence of filter settings on defogging effect and the cause of halo phenomenon after defogging are analyzed.Aiming at the problem that it is difficult to maintain the detail and color balance of Retinex algorithm and the halo problem caused by the change of brightness after defogging,a dual color space defogging algorithm based on Retinex theory is proposed,and the contrast component is estimated by bilateral filtering.Firstly,in HSV color space,higher-scale bilateral filtering is used to estimate illumination component on brightness channel,and adaptive Gamma correction is performed on saturation channel.Secondly,in RGB color space,the illumination component is estimated by lower-scale bilateral filtering on three color channels,and the reflection component is enhanced by S-function.Finally,the images after defogging in two color spaces are fused at the pixel level in RGB color space.The experimental results show that the improved algorithm has obvious details enhancement effect,high color fidelity and no halo phenomenon.In the field of image restoration,image defogging algorithm based on dark channel prior is mainly studied.The influence of atmospheric light on the results of defogging and the reason of low transmittance estimation in dark channel prior failure areas such as sky are analyzed.To solve the problem of inaccurate estimation of atmospheric light,the global atmospheric light with a single value is extended to the local atmospheric light map.The reference atmospheric light obtained by quadtree algorithm is combined with the bright channel image to generate atmospheric light map,which improves the accuracy of atmospheric light intensity estimation and makes it suitable for different regions of the image.Aiming at the problem of color distortion caused by low prior transmittance estimation of dark channel,a prior failure domain identification model of dark channel based on the characteristics of transmittance and bright channel image is proposed,and the region with low transmittance is compensated adaptively by using bright channel.The experimental results show that the scene is bright and clear after fog removal,and the deficiency of color distortion after fog removal in sky area is improved.