Single Image Defogging Research Based On Atmospheric Scattering Model And Segmentation Algorithm

Haze is a common weather phenomenon.In this kind of weather,the image quality is generally reduced,such as the contrast is reduced,the details are lost,which brings great difficulties to the image processing.In order to obtain high quality image in haze weather,it is necessary to defog the fog image,improve image contrast,enrich the details of the image,ensure the computer vision system can work in haze weather.There are two kinds of research on fog image defogging: one is the defogging method based on image enhancement,the other one is the defogging method based on physical model.Image enhancement methods enhance image details and improve the visual effect of the image.This kind of method brings some problems,such as color distortion,image distortion and so on.The defogging method based on physical model established the physical imaging model(such as the atmospheric scattering model)of the fog image by analyzing the causes of the image degradation,and realized the restoration of the fog image by inverse operation of the model.The good defogging effect is realized by analyzing the imaging process of fog image.At present,there are many defogging methods based on physical model,among them the defogging algorithm based on atmospheric scattering model and prior knowledge has achieved very good defogging result,but this method still has some problems.Because of the scope of image defogging algorithm being limited by prior knowledge,it lacks universal applicability.In order to solve these problems,this paper proposed a method to predict the scene transmission based on the fog related features.The most effective four fog related features are screened out by analyzing the fog related features in a large number of researches,and the established random forest model between them and the scene transmission have been elaborated studied.Finally,the clear image is obtained based on the atmospheric scattering model.In order to solve the problem of dark channel prior defogging algorithm in sky area and pure white area,this paper proposed an improved algorithm of dark channel defogging based on image segmentation.Firstly,the fog image is divided into sky and non sky region by machine learning method,then the non sky region is divided into pure white region and normal region by binary image fusion method.Different strategies are used to calculate or modify the scene transmission of these three areas.The transmission fusion method based on the color attenuation prior and the dark channel prior is used to calculate the transmission of the sky region and normal region,the transmission compensation method is used to calculate the transmission of the pure white region.Finally,the clear image is achieved by atmospheric scattering model.Through the comparison experiment,one can find that the image obtained by the defogging algorithm is more clear and natural,and it doesn't have obvious color distortion in this paper.Through the objective image quality assessment,we can further verify that this algorithm has achieved very good defogging effect.