Research On Image Dehazing Method Based On Physical Model

With the development of science and technology,computer vision applications such as remote sensing,unmanned driving,and target recognition have become more and more widespread.However,outdoor images acquired in foggy weather will be affected by fog,resulting in severe degradation of image quality.The foggy image has the characteristics of low contrast,decreased scene clarity,and dim color.Image quality is greatly reduced in foggy weather,which limits the work efficiency of image-based application systems.Therefore,the research of image defogging has very important practical significance.In the image defogging method based on the physical model,the image defogging algorithm based on the color attenuation prior theory is first studied.The influence of atmospheric light on the results of defogging and the causes of fog residue and blurry details are analyzed.In order to solve the problem that the estimation of atmospheric light is not accurate enough,the quadtree search algorithm is used to obtain atmospheric light.Aiming at the problems of blurry details and fog residue in the image,an improved method of image defogging based on a priori theory of color attenuation is proposed.A watershed algorithm based on spatial constraints is used to superpixel segmentation the scene depth map.After using the superpixel algorithm,each area will be labeled,and then combined with the filtering algorithm to process these labeled areas separately to obtain the scene depth information map.The experimental results show that the proposed improved method can effectively improve the problem of blurring details and residual fog in the image,with obvious enhancement of details and natural colors.In the image defogging algorithm based on the dark channel prior theory,the reason why the sky and other areas do not conform to the dark channel prior and the problem of insufficient accuracy of atmospheric light estimation are analyzed.In order to solve the problem that the atmospheric light estimation is not accurate enough,the quadtree search algorithm is also used to obtain the atmospheric light value.When the dark channel a priori defogging algorithm handles foggy images,problems such as color shift and distortion are prone to occur in the sky or large areas of bright areas.It is proposed to use K-means clustering algorithm to preprocess the image,mark the sky or a large area of bright white area,and then adjust the pixel intensity to solve the phenomenon of color distortion.In addition,the wavelet image fusion algorithm is used to refine the transmittance,so that the image detail information after defogging is enhanced and the halo phenomenon is improved.It has been verified by experiments that the improved method can refine the transmittance and at the same time effectively improve the problem of color distortion after defogging in the sky area.