Improvement Of Image Algorithm Based On Sky Recognition And Dark Primary Priori

With the development of global climate change and industry and urbanization,there are more and more haze.Fog and haze weather leads to low visibility,which brings inconvenience to people's daily life,aviation and navigation,and traffic supervision.It also leads to the loss of information in different of images taken in haze weather.In order to better restore the lost image information of fog image,it is very necessary to defog the fog image.This paper is of great significance to study the fog image algorithm.At present,there are two methods to remove fog effect from images at home and abroad: one is the method which based on image enhancement and another one is the method based on the atmospheric scattering physical model to remove the fog effect,this method has using the establishment of atmospheric scattering model to defog the image,because using the physical model could dig the lost information of the original image in depth,this method has become the most important tool to defog the image.So this thesis bases on analyzing physical model to defog and puts forward a advance way that combined the sky segmentation algorithm with dark grey prior algorithm,optimize the quality of defog image,so that good defog quality could be obtained.This thesis mainly does the following work:1.By consulting the documents,to understand the mainstream methods of defog image at home and abroad and the advantages and disadvantages of the methods are analyzed,and put forward the research object of this thesis.2.The physical principle of producing haze weather is described,and the atmospheric scattering model is explained in detail.This thesis also introduces the image enhancement algorithm,such as: the histogram uniform algorithm,the retinex algorithm.And the defogging algorithm based on the physical model,such as: the fattal algorithm and the tarel defog algorithm.Reproduce algorithm on computer and deal with foggy image,The principle of calculation and its advantages and disadvantages are analyzed.3.To have a detailed understanding of today's most popular dark grey prior algorithm and the meaning of each parameter in the atmospheric scattering model,and analyzes the dark grey prior principle couldn't apply in the region of the sky,these reasons leads to the phenomenon that block areas in the sky after image defogging.4.Based on the limitation that dark grey prior algorithm couldn't apply in the sky region.This thesis puts forward a advance way that combined the sky segmentation algorithm with dark grey prior algorithm,and then to learn the traditional sky segmentation algorithms,and analyzes its shortcomings,thus put forward the way of improving the sky recognition segmentation algorithm,and optimizing the transmittance t(x),and also optimize the atmospheric light value A,make theprocessing time of dark channel to defog algorithm declined dramatically.5.To have a objective evaluation of the defog images which using the advanced algorithm,this kind of defog image is better on peak signal.to.noise ratio,the mean square error(mse),the information entropy,average gradient,structural similarity than the traditional dark channel prior algorithm.