Research And Implementation Of Image Defogging Technology

With the rapid development of computer hardware and software,computer vision system has been widely used in intelligent traffic,video surveillance,target tracking and other fields.However,in hazy and foggy weather conditions,the images captured by the camera are seriously degraded due to the scattering of the suspended particles in the atmosphere.These images have low visibility,which directly limit and affect the normal operation of the computer vision system.Therefore,in order to ensure the normal work of visual system,it is very important to improve the visual effect of foggy images and reduce the effect of fog.In this paper,we detailed analyzed the image degradation physical causes and proposed a novel image dehazing method.The main research contents are as follows:(1)Under the problem that many dehazing methods can't get good dehazing effect in different hazy environment,we proposed an adaptive single image haze removal method based on physical model.For the important variable transmission in the dehazing model,we use the scattering coefficient and the depth of the scene to obtain.In the process of solving the scattering coefficient,we consider the diameter of haze particle as an important parameter.We can automatically identify haze levels by its dark channel image and determine the value of haze particle diameter.The depth of scene is estimated by maximizing a posterior probability and we can get good dehazing effect in different hazy environment.(2)As the inaccuracy of depth image estimation using the maximum a posterior probability method,we proposed a dehazing method based on binocular stereo matching.The accurate depth map is estimated by using stereo matching technique.We also use the fast guided filter to improve the speed and the dehazing effect is better.(3)We designed and implemented an image dehazing system based on our dehazing methods.Users can get the dehazing effect and some evaluation index data while adjusting parameters.