Research On Optimization Algorithm For Single Image Dehazing Based On DCP

The rapid development of society has provided a lot of convenience and benefits to people's lives.On the contrary,it has also caused some unavoidable environmental problems,such as bad weather.Haze has become a common weather phenomenon,which not only endangers people's health,but also causes problems such as low contrast,low definition and color shift in images collected by outdoor collection equipment.So image dehazing has become a hot research direction in the field of computer vision.The dark channel prior(DCP)dehazing algorithm based on the atmospheric scattering model has achieved good dehazing effect,but there are still many problems.Later,many researchers improved the algorithm and proposed many optimization algorithms.Until now,DCP has become the basis of many dehazing algorithm.This paper based on DCP is improved,and researching related improved algorithms is to obtain higger quality images.The key research contents of this paper are as follows:1.The improved single image dehazing algorithm based on DCP is proposed.According to the size of the image itself,an adaptive filtering window is proposed;the variogram is used to remove the influence of the highlight pixels on the atmospheric light value;the factor w is determined by combining with the structural similarity of the image;in order to present the better effect of dehazing,the brightness of the defogged image is improved by using the conversion between RGB model and HIS model.The experimental results show that the improved algorithm achieves better dehazing effect than the experimental comparison algorithm.2.The single image dehazing model based on the parameter driven and the improved algorithm are proposed.Firstly,inspired by the dark channel prior,the bright channel prior theory is derived,the dark channel prior theory is combined with the bright channel prior theory,and the transmittance is calculated adaptively based on linear weighting of parameters.Secondly,a parameter-driven dehazing model is proposed,which uses multiple selected objective evaluation indicators of image quality to determine the parameters corresponding to the adaptively obtained transmittance.Finally,on the basis of the the model,an improved parameter-driven dehazing algorithm is proposed.The experimental results show that the improved algorithm not only retains more detail information,but also has a good effect in dehazing from the objective evaluation index of image quality.3.The atmospheric light value estimation method based on the improved quadtree is proposed.The atmospheric light value is generally taken in the dense fog area of the sky.Based on this feature,the average pixel value in the window area is taken as the basis of quadtree segmentation,and the area with the largest average pixel value is continued to be segmented until the window is less than or equal to the adaptive window threshold.The adaptive window threshold can effectively prevent filtering out the sky area and avoid the interference of bright objects,making the estimated atmospheric light value more accurate.The improved algorithm is more exact than the classic estimation algorithm and the estimation algorithm in Chapter 3.The proposed algorithm consumes less time and the efficiency of dehazing is greatly improved.