Research On Video/Image Recovery Method Based On Harsh Environment

The research and application of computer vision and digital image processing have received more and more attention from people.The outdoor vision system has also achieved significant results in many fields,such as intelligent transportation,video surveillance,and remote sensing monitoring.However,the harsh environment causes extremely severe interference to the outdoor vision system.The typical strip noise present in the hyperspectral image and the raindrops and smoke common in the video can greatly damage the quality of the image or video.In order to obtain high-definition and high-quality images or videos,it is extremely important to eliminate the negative effects of these noises.This thesis mainly studies the degraded video or image restoration methods in the three kinds of harsh environments,such as the strip noise in the hyperspectral image and the rain and smoke appearing in the video.The main work is as follows:(1)Considering the low rank characteristics of clean hyperspectral data and the differences between different spectral images,this thesis proposes a tensor completion method based on spectral space consistency regularization to remove strip noise.Then,this thesis uses the converged multi-module Alternating Direction Method of Multipliers(ADMM)to solve the algorithm model.Finally,the experiment results are used to verify the effectiveness of the algorithm in this chapter,and the experiment results show that the proposed algorithm achieves good results both objectively and subjectively.(2)This thesis optimizes and improves the existing traditional algorithms from the smoothness and sparsity of raindrops and the low rank characteristics,local and global consistency of video.Considering the influence of wind,the rain is inclined to a certain angle with the horizontal direction.And then,the rotation operator is added to propose a new adaptive video rain removal algorithm model based on tensor completion.And then,using the alternating direction multiplier method to solve and finally get the experiment results.The experiment results show that the rain removal algorithm proposed by this thesis has obvious advantages compared with other algorithms.(3)Considering the difference between smoke and haze,this thesis proposes a low-rank tensor-based smoke removal algorithm from the perspective of space-time consistency.And then,the Alternating Direction Method of Multipliers is used to solve the proposed smoke removal algorithm model.The experiment results show that compared with other dehaze algorithms,the smoke removal algorithm has significant smoke removal effect on the problem of smoke treatment.