Research On Haze Removal Algorithm Of Foggy Degraded Image

Because of the influence of the scattering media in the atmosphere,foggy images will be degraded in foggy environment,which greatly affects the performance of outdoor vision system based on clear images.Since the weather is uncontrollable,the usual method is that use the subsequent digital signal processing to dehaze the degraded image,so that the image quality can be improved,thus improving the performance of the outdoor vision system.Therefore,the research of image dehazing algorithms in foggy degraded images has important theoretical value and practical application prospects.In this thesis,the foggy imaging model is applied to analyze the foggy image characteristics,and haze removal is done from two aspects of image restoration and image enhancement.In terms of image restoration,an image dehazing algorithm based on prior information is studied and implemented.In view of the inadequacy of dark channel prior and color attenuation prior in the accurate estimation of transmission,dark channel prior and color attenuation prior are combined by disassembling the foggy imaging model.The nearby/distant regions can be segmented by value-saturation difference information,so that different regions can be dehazed by different models,and reasonable restrictions on the values of atmospheric light are made,thereby reducing the problem of color distortion in the distant area.The image dehazing algorithm based on prior information studied in this thesis can better maintain the balance between the two aspects of color information and detail information recovery,and has good image haze removal effect.In terms of image enhancement,an image dehazing algorithm based on color constancy is studied and implemented.Aiming at the problem of color distortion and time-consuming in the application of the original Retinex theory in haze removal,Retinex theory is applied to two color spaces of RGB and HSV respectively.The color recovery coefficients are solved in the RGB color space,while the luminance components and saturation components are processed in HSV color space,so as to minimize the color distortion.Guided image filter is used to quickly solve the image reflection component,and different sizes of filter window are set up to realize the multi-scale Retinex algorithm.Color constancy based image dehazing algorithm studied in this thesis maintains a good balance between color naturalness and colorfulness,achieves detailed information recovery and has a good image dehazing effect.All the algorithms studied in this thesis are simulated,and the performance of the haze removal is evaluated by subjective-objective evaluation.Experimental results show that the studied and implemented algorithms outperform the classical image dehazing algorithms in controlling image distortion,improving image contrast,color naturalness and colorfulness,and recovering image details,and the algorithms have certain practical value.