Research On Model And Algorithm For Video Rain Streak Removal

Video rain streak removal algorithm has always been an important research content in the computer vision and image processing.Especially in reality,because the rain streaks will block the condition of the road to a certain extent,driving in heavy rainy is very dangerous for drivers and to a certain extent,it restricts the development of the artificial intelligence such as automatic driving.what is more,because of the influence of rain streaks,the camera in important positions will also be blocked,making the camera equipment unable to provide clear images.It has a bad impact on locating criminals for the police and the confirmation of responsibility in traffic accidents.In daily life,when people take pictures,the pictures may inevitably have the rain streaks which makes people can not get a clear image.So we need to do image process to the images to remove the rain steaks and ensure the clarity of the image.In this thesis,we propose a novel tensor optimization model for video rain streak removal by fully considering the discriminatively intrinsic characteristics of rain streaks and clean videos.In specific,rain streaks are group sparse and smooth along the rain streaks' direction.The clean videos are smooth along the perpendicular direction of rain streaks and the temporal direction.For rain streaks,we use the ?2,1 norm to enhance the group sparsity and the Unidirectional Total Variation(UTV)to promote the smoothness along rain streaks' direction.For clean videos,we use two UTV to enhance the smoothness along the perpendicular direction of rain streaks and the temporal direction.To solve the proposed tensor optimization model proposed,we develop an effective alternating direction method of multipliers(ADMM)to solve the proposed model.The proposed model is a large convex model,but it can not be solved directly in an expression.In order to solve the proposed model,we introduce four variables to decompose the coupled variables into two groups.This allows us to get the optimal solution of the variables of one group we fix the variables of another group.This conforms to the ADMM framework.Through the alternate iteration updates of the variables of the two group and the dual variables,we can get the clean video.At the same time,the framework of ADMM guarantees the convergence of our algorithm.Experiments on synthetic and real data demonstrate the superiority of the proposed method over state-of-the-art methods in terms of both quantitative and qualitative assessments.