Rrsearch On Image Degradation And Recovery Based On Physical Properties Of Atmosphere

The optical imaging technology plays an important role in the areas of transportation, agriculture, scientific research, and even in our daily life. However, the image quality is influenced by the interaction between the light and the particles in the atmosphere. Due to the existence of fog or haze, the contrast of the image declines and less information can be extract from the captured image. Especially for the remote sensing, the image degradation is dramatic because of the long transmission path. Therefore, the use of the imaging equipment is limited. Fortunately, it is possible to remove atmospheric influence from images using digital image processing methods, and the objects covered by the haze can be re-seen. The thesis focuses on the physical process of image degradation along the atmospheric path and the image recovery algorithm, which are summarized as follows.Firstly, we analyze the optical properties of the atmospheric particles, including the absorption and scattering. In the visible spectrum, the main absorbing gas is water vapor, carbon dioxide and ozone. The infrared-absorbing gas is water vapor and carbon dioxide. The atmospheric absorption causes the exponential attenuation of the radiance according to the Bouguer-Lambert-Beer law. However, the effect of the atmospheric scattering is more complicated. The molecule obeys the Rayleigh scattering theory because the size is much less than the wavelength, while Mie scattering theory is used for aerosols. A fraction of the radiance can be scattered out of the view-of-sight, and due to the multiple-scattering, a fraction of the leaving-radiance is able to entrance the view-of-sight again. According to the effects of absorption and multiple-scattering, we propose an image degradation model to explain the imaging process among the atmospheric medium, especially for the long distance and poor weather condition.Secondly, we study the image recovery methods, which can be divided into two classes, including multiple-image haze removal ones and single image haze removal ones. Depending on the demands for the efficiency or the accuracy of the method, we propose three image recovery algorithms based on our model, i.e., the fast single image dehazing algorithm using the existing values of the parameters which are obtained from the PcModWin software, the single image dehazing algorithm using blind deconvolution method in order to acquire the atmospheric point spread function and the image dehazing algorithm based on the image fusion method using one visible image and one infrared image.At last, we verify the effectiveness of our model and algorithms by simulation experiments and real captured image experiments. We also discuss the relationship between the error amount and the dehazing result during the simulation experiments. In order to compare the performances of the recovered images objectively, several image quality assessment methods are used. The results obtained from the proposed model and algorithms indicate better effect.