Image Haze Removal Algorithm Based On Filter And Physical Model

Haze appears frequently and is widely distributed,it is a common weather phenomenon in China.Due to such bad weather,it is inconvenient for people to go out and greatly increased the incidence of traffic accidents.Affected by haze,Imaging equipments can not identify or extract image features at the same time.Therefore,research on image haze removal has been particularly significant and valuable.At present,the methods of image haze removal are classified into two categories: image enhancement method and image restoration based on physical model.This paper makes a deep research on several mainstream algorithms among the two categories,the main research works are as follows:(1)Image enhancement method has been studied in full detail.The research mainly focus on the histogram equalization algorithm,homomorphic filter algorithm and Retinex algorithm.Through a large number of simulation experiments to analyze the advantages and disadvantages of these algorithms,a new algorithm is proposed: Homomorphic filter suppresses the irradiated components.Retinex filters out the components of the irradiation.Both of them deal with foggy images based on illuminance-reflectance model,try to remove the irradiated light and retain the reflected light.Therefore,the new algorithm first performs improved homomorphic filtering on the RGB channels of foggy images,suppresses the irradiated components while compressing the dynamic range of images and smoothing the light.Then,each channel is dehazed by Retinex.Finally,the images of each channel are combined to obtain the final result.Experimental results demonstrate that new algorithm can obtain better effects.(2)Image restoration based on physical model has been studied in full detail.The research mainly focus on single image haze removal using dark channel prior and several improved algorithms.In order to solve the problem in image haze removal algorithms based on dark channel prior like color distortion and blurred image edge information,an improved image haze removal algorithm based on blockwise processing is proposed.Firstly,LAB color space is chosen instead of RGB color space.Then,bilateral filtering is used to preserve edge smoothly and refine the transmission map.Finally,a tolerance is set for blockwise processing to remove color distortion problem.The results demonstrate that,the new algorithm can successfully solve the problems like color distortion and blurred image edge information.(3)Mean value,standard deviation,information entropy and average gradient are selected as standard objective assessment to evaluate dehazed image from different aspects.Finally,this paper summarizes research works,and prospects for follow-up work.