Research Of Image Defogging Algorithm Based On The Polarization Characteristics

Bad weather, such as fog and haze, can significantly degrade the visibility of ascene. It is always a hot topic in how to efficiently remove the influence of the hazeand fog on the optical images, or how to improve image quality and get usefulinformation which users are interested in. In recent decades, image restorationalgorithm based on polarization property has attracted much attention from manyresearchers. The algorithm can be widely used in fields such as intelligent trafficmonitoring and military reconnaissance, etc.In this thesis, an algorithm for defogging of polarimetric images is developed.Further, the algorithm is applied to road monitoring system. This thesis consiststhree parts as follows.(1) Relationships between atmospheric light scattering and scene reflection aswell as connections between atmospheric polarization and the optical thickness inhaze weather are investigated. Traditional polarization image defogging algorithmhas a drawback that degree of polarization is approximately considered as a globalinvariant. Based on the principle of dark grey method, the algorithm estimating thedegree of polarization accurately is proposed in this thesis.(2)Based on the analysis of classic polarization image defogging algorithm, animage restoration algorithm is developed, which based on polarizationcharacteristics. First, a polarization imaging model is established. Neighborhoodaverage intensity values of qualified pixels are used to estimate the atmosphericlight intensity. Moreover, the information of depth of field is applied to refine theglobal degree of polarization to the local. Finally, defogging of images is carriedout. Compared with classic algorithm, the algorithm presented in this thesisachieves higher robustness and gets images with higher contrast and clearer details.(3)Based on the polarization images defogging algorithm, the thesis designsand implements an electronic police monitor system, which can deal with fogimages captured under the condition of fog and haze, so as to provide clear imagesfor license plate recognition and driver identification. The result of actualmeasurement shows that the polarization images defogging algorithm is effectiveand feasible for practical application.