Imaging Through Turbid Media Utilizing Polarimatric Information Of Light

In some special weather conditions,like rain,fog,and haze,large amount of particles including water drops and aerosols suspend in the air,resulting in sharp attenuation of light while propagation.Images taken under these conditions suffer from serious degeneration in contrast,details of objects,and so on.Fog,haze and water drops in the air are labeled as turbid media in this study,and we aim at recovering clear images from images taken through any of these turbid media.In this study,we proceed from two most common weather conditions: haze and water spray to figure out the effects of turbid media on imaging.The study starts by analysis of the atmospheric scattering model of light propagating in the turbid media,with emphasis on researching the polarization characteristics of scattered light.In accompany with the scattering model,polarization imaging system is designed and applied to capture images on both hazy and rainy days.Specified algorithms are designed to match the polarization images to finally recover clear images.There are four important parts introduced and completed in this dissertation.1.We established the image degradation model under hazy and rainy weather based on the analysis of light propagation model and scattering model.For obtaining clear images in hazy conditions,capture of the light’s polarized information was taken as the key step.We first set up a polarization imaging system to capture images.Then the frequency spectrum was generated and analyzed.On the above basis,a multi-scale dehazing algorithm was applied to remove the effects of haze on the images.Experiment results on both mist and dense haze conditions demonstrated the satisfactory results in recovering clear images.2.When processing images taken under dense haze,the dehazed images presented color distortion in some parts with high degree of polarization.To solve this problem,the modified multi-scale polarization dehazing algorithm is established based on the polarization information of both airlight and direct transmission.It can handle the color distortion in dehazed images satisfactorily.3.Special attention was paid to images taken in rainy conditions.Different from other particles in the air,water drops could change unpolarized incident light into partially circularly polarized ligtht and keep the circular polarized state unchanged.This special signature enabled imaging in rainy conditions by a different way.A novel real-time circular polarization imaging system has been designed and optimized based on the Wollastopn Prism in order to capture two orthogonally polarized images simultaneously.Correspondingly,a water spray removal algorithm based on the circular polarization information was conducted.Experiment results proved the algorithm could enhance image contrast and increase detection range.4.With deeper research of the scattering model of light propagating in turbid media and the atmosphere attenuation model,we took actions to calculate the depth information of the target scene.The key step lay in that during imaging,two images with different polarization information were captured.The depth of targets in the scene could be calculated by the scattering model employing polarization information.Experiment results showed that the scene depth information model utilizing the polarization imaging method could reconstruct the depth image and calculate the relative distance of the object.