Research Of Fog-degraded Image Restoration Method Based On The Atmospheric Scattering Model

In bad weather, such as fog and haze, images acquired by imaging device are degraded severely by scattering due to floating particles in the atmosphere, showing poor contrast, color distortion, low clarity and other characteristics. But image-related processing algorithms in existing outdoor vision systems are only suitable for the image acquired under sunny weather conditions, this results the outdoor visual system can’t work properly in bad weather. In order to ensure outdoor vision systems can be more robust and reliable, doing fast and efficient image clarification for foggy image has important significance.The dissertation researched on image clarification from the point of image restoration, proposed and implemented a fast and efficient algorithm for image clarification. Implemented experiments on several outdoor foggy images, and the experimental results were analyzed and evaluated in detail combining subjective observation and objective evaluation. The main contents are as follows:1) The dissertation described the research background and current status of technology of image clarification. Deep researched on the basic theories and key technologies of image restoration. Then detailed described the effects of atmospheric scattering for imaging device, and established corresponding degradation model. According to the degradation model, the nature reason of image degradation and features of foggy image were analyzed.2) For the deficiency of fast defogging based on the atmospheric veil, a fast foggy image restoration algorithm based on high precision atmospheric veil was proposed. Firstly, from the atmospheric scattering model, we introduced atmospheric veil in the algorithm to simplify the model. Then a precise method for solving the atmospheric was put forward based on the thought of finding the sky part or the most hazed region of the image. Finally, we got the high precision atmospheric veil by calculating the PauTa criterion limit based on the idea of class morphology. After all the above, the fast restoration of foggy image was realized according to the simplified atmospheric scattering model. The proposed algorithm can exactly restore the color and clarity of the scene, and the complexity of the proposed algorithm is just a linear function of the number of input image pixels, this allowed a great promotion in the computing speed.3) Defogging for single foggy image based on dark channel prior has achieved great results, but this method expends large amounts of memory storage and computing resources, results it difficult to apply. Besides, the result of transmission is not accurate enough when the dark channel prior is failure. For the deficiency of defogging used dark channel prior algorithm, a series of improvements were proposed. First of all, combined image down sampling and up sampling methods, we still use soft matting algorithm to refine the transmission obtained by dark channel. Then, introduced adjustment factor to recalculated transmission in failure dark channel prior area, such method can avoid color distortion effectively. At last, stretched the brightness adaptively for restored image. Experiments show the improved algorithm can quickly eliminate the fog and haze in foggy image, restore the true nature of original scene, and the enhancement of speed make real-time application of defogging algorithm be feasible.