Research Of Polarized Image Dehazing Technique Based On Color Space Conversion

In recent years, frequently appeared fog and haze have caused attention widely. Fog and haze not only affect our healthy, but also greatly influences on transport and intelligence operations, as well as the effect of a lot of intelligent monitoring systems and detection equipment. For example, we can’t aim missiles at targets on the battlefield, can’t identify the license plate traffic monitoring in the road, and can’t track objects of the investigation on-site equipment, etc. So how to remove fog and haze has become one of the hottest spots in the image visual processing research.There are a variety of imaging ways, including common visible light, remote and polarized sensing, infrared and so on. Compared with visible light and infrared imaging system, polarization imaging has many advantages, it can get multi-dimensional radiation polarization information feature, which can improve the ability to identify the target. Effectively using polarization information can get more image details and can improve the Signal to Noise Ratio (SNR) of the image. Most traditional image defogging algorithms based on polarization characteristics are focus on optimization of atmospheric light and depth information. In high visibility condition it can effectively remove the image. On the contrary, the effect is not ideal. Due to the restoration result under low visibility conditions is not satisfactory, we propose a new color space conversion algorithm using dark channel prior for polarization image dehazing in the thesis. The main work is as follows:Firstly, we discuss the haze formation mechanism and atmospheric scattering, combined with atmospheric scattering model, study the image defogging model based on dark channel priori.Secondly, we study the atmospheric polarization characteristics, research the traditional image defogging algorithm based on polarization characteristics, and analysis the shortage of traditional polarization image defogging algorithm. Then we propose a new color space conversion algorithm for polarization image defogging. Polarization images usually use intensity, polarization degree and polarization angel information to describe target’s polarization information, in order to combine the polarization information and defogging model, convert the polarization information into HIS color space.Thirdly, we revalue the chromaticity and saturation information of HIS color space, map the HIS color space to the RGB space, and then elaborate the atmospheric transmission rate using a guide filter algorithm based on dark channel prior.Fourthly, we compare the experimental results with the traditional polarization characteristics image restoration results. The result shows that many technical specifications of defogged images such as standard deviation, entropy, average gradient will be improved greatly in low visibility conditions, and our method can effectively improve the identification capability for the polarization image.