Research On Definition Processing Of Haze Image Based On Binocular Stereo Matching

With the sustained economic growth and the rapid development of automobile industry and industry in China,the content of dust particles in the atmosphere is increasing.Especially in winter,it causes different degrees of air turbidity,which is what we usually call fog and haze weather.Because of the high humidity and water vapor in haze weather,droplets provide adsorption and reaction sites to accelerate the conversion of reactive gaseous pollutants to liquid particulate matter,and particulate matter is easy to accelerate the formation of haze as condensation nodules.On the one hand,in severe fog and haze weather,highway traffic restrictions,aircraft flight delays and grounding often occur,resulting in a certain degree of restrictions on people's daily travel;on the other hand,fog and haze weather due to the presence of a large number of suspended particles,narrowing people's sight range,blurring the visual effect,while resulting in video surveillance image quality decline and The contrast becomes worse,which brings great potential safety hazards to people's personal and property.At present,the processing of haze image at home and abroad only stays at the stage of defogging.Because of the different components of fog and haze,the optical imaging principle is also different,so the defogging algorithm can not deal with the haze image very well.The main work of this paper is as follows:(1)The basic principles of histogram equalization algorithm,dark channel priori algorithm and Retinex algorithm are studied.The simulation experiments of each algorithm are carried out,and the experimental results are presented and analyzed.(2)In order to avoid the influence of weather and region on the collection of haze image,a binocular clear image haze algorithm based on depth map is proposed in this paper,which provides data for the further study of binocular haze image sharpening algorithm in this paper.(3)Aiming at the characteristics of low contrast,clarity and brightness of haze image,a binocular stereo matching haze image sharpening algorithm is proposed,which is more suitable for haze image processing,combining the imaging principle of haze,scattering distribution principle and the characteristics of scene depth information of haze image.Firstly,the algorithm decomposes the image taken in haze environment into the radiation component of the scene(clear image without haze)and the scattering component caused by suspended particles(the part of haze image to be removed);secondly,researchers have known that all the haze images taken at infinite distances are composed of scattering components,therefore,the scattering component and distance.The distance from the target scene is related to the depth of field,and there is a functional relationship between them.In order to find the scattering component at any point in the target scene,a scattering proportional function is defined.This proportional function is used to describe the functional relationship between the scattering component and the depth of field.That is,the scattering ratio is the ratio of the scattering component of the target scene to the scattering component at infinite distance.Furthermore,it can be deduced that the scattering component at any point in the target scene can be expressed by the scattering scale function and the scattering component at infinity.Then,the region-based stereo matching is carried out by using the normalized sum of squares of deviations.Only when the scattering scale function is minimum,the binocular clear image containing only the radiation component of the target scene can be obtained.Finally,the SIFT algorithm is used.The method searches for the key points of the binocular clear image and then stitches the image to get a complete clear image.(4)Based on the research of binocular haze image sharpening algorithm,a prototype system of binocular haze image sharpening processing is programmed and implemented.The system mainly includes binocular haze image reading module,haze image sharpening processing module and clear image output display and preservation module.Based on the difference between haze and haze components and the optical imaging principle of haze images(atmospheric scattering model and atmospheric attenuation model),a new haze processing algorithm for binocular clear images and a binocular stereo matching sharpening algorithm for haze images are proposed in this paper.The data source of this paper is on the one hand to use Fuji 3D W3 binocular stereo camera to capture the collected binocular stereo haze image on the spot,on the other hand,to use the proposed binocular clear image haze algorithm based on depth map to synthesize the binocular stereo haze image.The simulation results show that the proposed algorithm achieves good results in both clear image haze and clear image haze.Therefore,the research results of clear image haze algorithm in this paper make the collection of haze images no longer affected by geographical,temporal and weather conditions,and provide convenience for shooting special effects of haze scenes such as 3D movies and TV series.At the same time,the research results of binocular haze image sharpening algorithm in this paper provide better processing of haze images(especially those with more haze components).It provides a new technical means.