Research On Image Dehazing Algorithm Based On The Image Statistical Properties

With the development of technology, the intelligence automatic system becomes moremature, outdoor monitoring system has been a very good development,too. But, because of thelow visibility, a large areas of image which is got by the outdoor mointoring system degeneratein the situation of rain、fog or haze and so on. Evently,the function of system will be reduced. Sothe research of dehazing algorithms which is about nature scene image becomes the hot point ofcomputer vision and image processing field. The image dehazing algorithm has a high practicalvalue to enhance the image contrast and improve the image quality.In order to realize the clearness of fog image, we improve the robustness of outdoor visionalsystems in this paper. The important research is dahazing algorithm which is based on the imagestatistical properties. The main researches of this algorithm are fog image degradation model,parameter estimation, image noise prior model and so on. The paper’s works are:First of all, we summarize previous image haze removal research, expound the significanceof study fog image restoration, introduce some basic image restoration knowledge. At the sametime, we describe the fog image degradation mechanism, analysis factors of image degradationmainly. These lay the foundation for the next work.Secondly, the fog image restoration has the uncertainties and contains multiple parameters.Aimed to these problems, we use dark channel prior method and area segmentation method toestimate parameters of the model of fog image degradation. In order to keep the robustness ofimage range, we propose a new estimate method of media function. The algorithm is mainly toestimate the median function in the use of the media filtering method. So that, the media functionbecomes more accurate. Experimental results show that the proposed algorithm can make theimage range more stable, save operation time of the program.Thirdly, we analyse three methods of image restoration under the bayesian framework. Onthis basis, we build a new degradation model according to the fog image degradation mechanism,which reduce the number of parameters, avoid the problem of the parameters estimation, and theless valuation,the more accurate of the final result. Fog image degradation model within thebayesian framework change the problem of image dehazing to a problem of optimization, reducethe difficulty of the problem solve.Fourthly, according to the sparsity of the image gradient, we build the prior models of fogimage and noise, and combining the proposed method of parameters estimate and new degradation model, we use the iterated weighted least squares to realize the fog image restorationbased on the bayesian framework. Experimental results show that the proposed algorithmremove haze effectively, improve the image quality, and keep the image robustness.Finally, we summarize all of this paper’s research content,and do a expectation for the futurework combined with the lack of work.