Image (De)Weathering For Road Based On Atmospheric Scattering Model

Due to the scattering of atmospheric particles, images of roads captured by visual surveillance system suffer from poor contrasts and colors under fog and haze weather conditions. This could frequently bring many difficulties in many aspects, such as road monitoring, target tracking, automatic navigation and so on. So it has a great significance to study the image clearness technique in these weather conditions.Based on the analysis of image degradation mechanisms in fog and haze, this paper conducts a full investigation into the restoration of degraded images from the point of atmospheric scattering model. The major researches of this paper are as follow: the first is to propose a new algorithm based on atmospheric scattering model for the degraded images de-fogged automatically without any ancillary information; another work is to carry out physical verification on embedded system for new algorithms, and optimize it on embedded system.The main contributions and creative works of the thesis are summarized as follows:(1) The atmospheric scattering model is simplified by reducing the number of unknown parameters to one.(2) An automatic method based on Hough transform is proposed to select the vanishing point.(3) The evaluation criterion of the quality of restored images is proposed by using entropy based on the theory of information. And the maximum entropy theory is used to acquire the optimal parameter.(4) Using the Davinci development platform of TI, the physical verification of new algorithm of image restoration based on atmospheric scattering model is carried out.(5) On hardware platform, the new algorithm has optimized, which have raised the speed of new algorithm greatly.Theoretical analysis and experimental results on real images demonstrate that, methods used in this paper are effective for clearness of fog-degraded images, and can be applied to the practical situations. Also, this research will be a inspiration for the further investigations on the process of images taken in other bad weather.