Research On Single Image Dehazing Algorithm Based On Adaptive Sky Region Segmentation

With the rapid development of intelligent technology and informatization,digital image processing technology has been widely used in unmanned driving,road monitoring,video security and other fields.Image processing systems often need clear images or image sequences to ensure the reliability and stability of the system.However,in haze weather,cameras working outdoors are vulnerable to the interference of a large number of aerosols in the air,thus reducing the amount of radiation from the scenic spots to the camera through the haze,and causing problems such as scene blurring and reduced contrast,etc.Therefore,it is of great practical significance to study how to effectively reconstruct the original haze free image from the hazy degraded image to maximize the restoration of details and texture information in the image.The main work of this paper is to improve the problem of image degradation by optimizing the traditional method based on dark channel prior on the basis of analyzing and studying the characteristics of a single hazy image,aiming at the problems of color distortion,halo and error estimation of atmospheric light value when some existing advanced algorithms such as dark channel prior algorithm process the sky region in outdoor images,so as to generate a high-quality clear haze-free image.The main research contents of this paper are summarized as follows:1.Improving the estimation method of atmospheric light parameters.The algorithm fully exploits the relationship between the characteristics of haze image in terms of brightness,saturation,smoothness,dark channel,etc.and the scene depth distribution,uses Canny edge detection method to obtain the rough edge line between the sky region and the non-sky region,and then uses morphology closed operation and the most general filtering algorithm to obtain the complete edge line and roughly segment the sky region in the image,so as to estimate the atmospheric light value from the sky region.2.Improving the estimation method of sky region transmittance.Then,based on the obtained edge line and atmospheric light value,an adaptive dual-threshold segmentation method is designed for the fine segmentation of the sky region,and a transmittance estimation method based on the light channel prior model is proposed to fuse the sky region transmittance value estimated by the light channel prior model with the foreground region transmittance value estimated by the dark channel prior model.3.Refining the fused transmittance and restoring haze free images.A fast guided filter is introduced to refine the edge of the fused transmittance,and the image restoration is completed based on the atmospheric scattering model.The experimental results show that the algorithm overcomes the defect of the dark channel prior algorithm in processing the sky region,retains the visual authenticity of the restored image and effectively removes the haze.