Research Of Image And Video Dehazing

Image and video dehazing is a crucial part of Computer Vision,which is widely applied in the Video Surveillance,Object Tracking,Ocean Exploring and many related fields.Recently,owing to the huge demand of image and video dehazing technology in practical problem,more and more researchers take part in the study of these topics,and propose many novel algorithms,which highly boost the performance of image and video dehazing.However,In order to boost higher performance,there are still some problems to be solved,such as color distortion in sky regions,local oversaturation,color inconsistency between the adjacent frames of videos and underwater video dehazing.Two main contributions of this article about image and video dehazing technology are listed below:1)To solve the inaccurate transmission estimation of sky regions and the inaccurate background light estimation in a single image or a video sequence,the algorithm of image and video dehazing using view-based cluster segmentation is proposed.Firstly,because the depth of sky region is much larger than the depth of nonsky region,so GMM is utilized to cluster the depth map to estimate the sky region and then the transmission estimation of sky region is modified to reduce color distortion.Secondly,this algorithm presents to use GMM based on Color Attenuation Prior to divide a single hazy image into K classifications and estimate the atmospheric light of each cluster respectively.Another,according to the correlation among the clusters,the atmospheric light estimation of every cluster is refined to improve global contrast.Finally,online GMM cluster is applied to video dehazing to reduce the computational complexity.2)To solve the underwater video dehazing,color inconsistency between the adjacent frames of videos and the influence of camera and object motions and water flowing,the algorithm of underwater video dehazing based on spatial–temporal information fusion is proposed.Firstly,according to the spatial–temporal information fusion between the adjacent frames of videos and the feature of the guided filter,the transmission estimation of previous frame and the gray image of current frame are applied to obtain the transmission estimation of current frame,and the least square method are used to reduce the influence of camera and object motions and water flowing to get more accurate transmission estimation.At last,due to the correlation of background light between the adjacent frames of video,the background light estimation of frames is modified to keep the color consistency between the adjacent frames of video.To the above two algorithms,the performance of them is evaluated by the qualitative and quantitative comparisons with other state-of-art algorithms respectively.Extensive experimental results demonstrate the algorithms can not only have superior haze removing capabilities subjectively,but also have better performance objectively.