Study Of Full-polarization Imaging Camera And Polarimetric Dehazing Technique

According to electromagnetic theory,four basic properties are included in light.These properties are amplitude,frequency,phase and polarization state,and each of them contains important information of the optical light.Optical imaging systems normally can obtain information of intensity(corresponding to amplitude),spectrum(corresponding to frequency)and spatial state(corresponding to phase),and then calculate the characteristics of the object.Similar to these properties,polarization state is also an important property,and contains abundant and unique information.The polarization property of the reflected light from objects is closely related to the surface roughness,geometrical shape,texture and structure of the objects.It may have similar reflectivity of different kinds of objects,but the polarization state will be very different from each other.Therefore,polarimetric imaging technique is a novel imaging technique based on detecting the polarization information of the reflected light from objects.By capturing several intensity images of different polarization angle,polarization information can be point-to-point calculated from the corresponding pixels of these images;and then substitute the polarization information for the gray level;and finally images with polarization information of objects can be obtained,including degree of polarization images,angle of polarization images and ellipticity of polarization images,etc.Compared with the traditional imaging techniques,polarimetric imaging technique has advantage in highlighting the objects.Polarimetric dehazing technique is one of the applications of polarimetric imaging technique,and has drawn wide attentions in recent years.Theoretical and experimental research on development of polarimetric imaging camera and kinds of new polarimetric dehazing algorithm has been studied in this dissertation.The main research contents and innovative achievements can be listed as follow:(1)In the task of full-polarization,simultaneous detecting,chromatic polarimetric imaging camera development,some work has been carried out.First,design a three-step calibration method for the polarimetric imaging camera.The first one is image registration.Ultilizing affine transformation model to registrate the images captured by the camera in real time,and the error is less than 2 pixels.The second is ununiformity correction.The camera is corrected by using the uniform non-polarized light emitting from the integrating sphere,and effectively restrain the vignetting effect of the lens.The third one is the calibration of polarimetric detecting precision.The main error is caused by detecting error of different transmittance from different polarizers and optical axis assembly error of polarizers.After the process of calibration,the linear polarizition detecting presicion of the developed polarimetric imaging camera can reach 95%.Secondly,develop two functions for the camera,which is full state of polarization imaging and polarimetric dehazing imaging,and perform ground probation.(2)The deep analysis on existing polarimetric dehazing algorithm shows that there are several drawbacks,such as poor real-time capability,requirement of sky region in hazy images and inaccuracy on estimating polarized parameters.Focusing on these drawbacks,a kind of novel polarimetric dehazing algorithm based on angle of polarization(AoP)is proposed.Innovations of this algorithm are that: first,utilizing Stokes-vector polarimetric imaging scheme,images with different polarization state can be simultaneously obtained,and this can solve the drawback of poor real-time capability;second,new algorithm on estimate the intensity of atmospheric light is proposed to avoid the constraint of sky region;third,the estimation accuracy can be improved by introducing AoP instead of degree of polarization(DoP)in the algorithm.Experimental results show that this kind of algorithm can effectively enhance the quality of hazy images.(3)For the case of dense haze condition,optimization of the proposed polarimetric dehazing algorithm is studied.First,restrain the noise brought by the detector using neighborhood averaging filtering method,improving the accuracy of detection of AoP;second,by analyzing the difference between dense haze and normal haze,the estimation of the airlight is developed to be adaptive to the dense haze.Based on this optimized algorithm,the quantitative experimental researches are explored.The experimental results demonstrate that,in dense haze condition,the contrast can be enhanced to 100%;the visibility can be promoted to about 70%.Besides,the dehazing algorithm can also be used in the smoke,dust and underwater condition.(4)In order to further promote the visibility of the proposed polarimetric dehazing algorithm,by combining fusing dehazing technique,a novel polarimetric dehazing algorithm is developed based on visible and infrared image fusion.It is a two-step algorithm: first,proceed the polarimetric dehazing algorithm to both visible hazy images and infrared hazy images,respectively,and obtain two dehazing images of each;then fuse these two images using multiscale directional nonlocal means filter,and obtain the final dehazed image.This kind of algorithm combines the advantages of polarimetric dehazing method and the fusing dehazing method,and therefore obtains dehazing images with better quality.Experimental results show that this algorithm has strong robustness and high reliability.In dense haze condition,the visibility can be promoted to over 100%.