Research On Image Haze Removal Algorithm

In recent years,haze weather has become more frequent,which not only affects people's daily life,but also limits the number of outdoor imaging systems.In the case of hazy weather,particles in the air absorb and scatter light,causing the images collected by these imaging devices to have some drop in contrast and fade in color.Therefore,for remote sensing,video surveillance,target tracking,and daily photography,which require higher-quality image behavior,the study of image dehazing algorithms has far-reaching significance.On the other hand,with the continuous improvement of computer technology,it also promotes image defogging technology to continuously enhance and achieve better results.This paper mainly studies the image defogging algorithm through the following aspects:1)Researched the general situation of the dehazing algorithm technology at home and abroad,mainly divided into two categories: image enhancement and physical model based.There are algorithms based on image enhancement,such as Retinex algorithm and histogram equalization.At the same time,based on the physical model,there are mainly methods such as using prior knowledge and depth of field estimation.Through a lot of research on them,the algorithms are simulated and their advantages and disadvantages are analyzed.Through the study of these work,a solid foundation for the in-depth study of subsequent algorithms is made.2)The basic theory of fog removal algorithm for fog images was studied,including the formation of fog,the atmospheric scattering model and the imaging principle.At the same time,the existing two kinds of classical algorithms are studied in depth,their effects are analyzed,and their deficiencies are summarized.3)An image dehazing algorithm is proposed for image degrading in misty conditions and using mathematical morphology and guided filtering.This method starts with the atmospheric scattering model and improves the existing method for obtaining atmospheric light values,thereby simplifying Atmospheric scattering model;secondly,the method of using mathematical morphology and guided filtering together with the simplified atmospheric scattering model to solve the transmission method is proposed;finally,based on the above two improvements,the image restoration simulation is performed and the existing representative defogging algorithm is performed.The contrast reflects the superiority of the algorithm.4)Aiming at the defects of some related manual features of traditional dehazing algorithms,using the advantages of convolutional neural network architecture,a new convolutional neural network architecture is designed and proposed for image defogging;finally,the simulation contrasts reflect this Compared with the existing traditional defogging algorithm,the algorithm is more abundant in reflecting details.