Image Dehazing Algorithm Based On Two Prior Knowledge

In bad weather conditions,especially fog,haze and other special weather make the image contrast decreased,the sharpness decreased,seriously affecting the quality of the captured image.It has become an important obstacle in the field of video surveillance,traffic monitoring,target tracking and identification.Therefore,image defogging plays an important role in the application of outdoor visual systems under hazy weather.The thesis studies the atmospheric scattering model under fog and haze weather,focusing on the analysis of the principles and key technologies of the dark channel priori,color attenuation priori and its dehazing algorithm.On this basis,two new defogging algorithms are proposed to achieve the recovery of fog images and obtain good results.The innovative work of the thesis is as follows:1)A Sky-distinguish dehazing algorithm based on dark channel prior is proposed.The algorithm improves the dark channel prior defogging algorithm from three aspects:sky region processing,dark channel map optimization,and transmission refinement.Firstly,Gaussian filtering and region growing algorithm are used to divide the image into sky region and non-sky region.Secondly,To improve the transmittance of the sky area by using tolerance mechanism and minimum threshold,effectively avoiding color distortion and disorder in the sky area;Thirdly,The dark channel image in the non-sky region is optimized using adaptive minimum filtering and segmentation mapping,and the lower limit of transmittance is dynamically adjusted to improve the image brightness.Finally,the defogged images of the sky and non-sky parts are merged to get the final defogging result.Experimental results show that the sky region segmented in this paper is accurate and the sky area of the restored image is smooth.It not only effectively solves the blockiness and color disorder in the sky,but also solves the problem of darkness and unclear edges in non-sky areas.2)An adaptive defogging algorithm based on color attenuation is proposed.The algorithm uses the color attenuation priori to establish a linear model of the transmission with respect to image brightness and saturation.Furthermore,by introducing additional visible edge ratio as the evaluation criterion,an adaptive method for obtaining the model parameters is proposed,then a set of adaptive algorithm for defogging is formed,and then the transmission map is optimized by the guided filter to recover the final haze-free images.The algorithm uses the evaluation index as the standard of the optimal parameters and the optimal recovery results without manual intervention.The experimental results show that the adaptive defogging algorithm in this paper can make the image sharpness and contrast improved greatly,and has advantage in real-time and clarity comparing with the other three algorithms.