Research On Single Image Dehazing Algorithm

In fog and haze weather,the scattering and absorption of particles such as water vapor,smoke,and micro water droplets in the atmosphere are serious,which makes the captured image overall blurred,low contrast,and color degradation.This brings difficulties to the application of subsequent images in the fields of target recognition,video surveillance,remote sensing aerial photography and so on.Studying the image dehazing algorithm to solve the problem of poor imaging quality in haze weather has important application value.The dissertation studies the atmospheric scattering model,analyzes the effect of haze on the imaging process and the causes of image degradation,and focuses on the principles and characteristics of the dark channel prior defogging algorithm,Retinex algorithm and image sparse representation model.The dissertation proposes two new defogging algorithms have been implemented to realize the clear processing of foggy images and improve the clarity of the images.The innovative work of the dissertation is as follows:(1)This dissertation proposes a single image defogging algorithm based on three-region division.The algorithm divides the image into sky,non-sky and transition regions.The sky region uses the adaptive large-scale Retinex algorithm to remove fog,the non-sky regions uses dark channel a priori defogging,and the transition regions uses a fusion method of two defogging results.The experimental results show that the proposed algorithm can solve the problems of color disorder and light spot in the sky area,and the sky is clear and smooth after fog removal.The false edge problem is overcome in the transitional region while maintaining image details.The non-sky area has played the advantage of the defogging of the dark channel prior algorithm and achieved good results.From the comparison results with different algorithms,the defogging effect of the algorithm in this dissertation is better than that of common algorithms as a whole.(2)This dissertation proposes an image defogging algorithm based on sparse representation.The algorithm thorough improves the image effect by increasing the image saturation and brightness to achieve the purpose of defogging.Firstly,the algorithm transforms image from RGB space to HSI space,and uses two-level wavelet transform extract features of image brightness components.Then uses the K-SVD algorithm training dictionary,learns the sparse features of the fog free image to reconstructed I-components of the fog image.Non-linear stretch saturation component to improve image saturation.Finally,convert from HSI space to RGB color space to get the defog image.Experimental results show that the algorithm can effectively improve the contrast and visual effect of the image.Compared with several common defog algorithms,the percentage of image saturation pixels are better than the comparison algorithm.