Research On Wmsn Video Denoising Based On Improved Model Of Low-rank Decomposition

Wireless Multimedia Sensor Networks(WMSN)is a distributed wireless communication network,it has a lot of advantages,such as flexible network layout,numerous types of transmission data and low cost of network construction.Therefore,WMSN has been widely used in video signal acquisition and transmission.However,the weather conditions,illumination intensity and video sensors own electronic characteristics lead to the result that WMSN videos inevitably suffer from various kinds of noise in the collecting and transmission of WMSN videos.Gaussian noise and rain streak noise are usually found in WMSN videos,which are damages to the WMSN videos,and had a negative impact on the further processing and analysis of the WMSN videos.Therefore,it is important to study the denoising methods for removing Gaussian noise and rain streak noise in WMSN videos.Firstly,this paper analyzes the basic principle,model,algorithm and its related application in the processing of low-rank sparse matrix.In order to solve the Low-Rank Sparse Matrix Decomposition(LRSMD)model,the Iterative Threshold(IT)algorithm,the Accelerated Proximal Gradient(APG)algorithm,the Augmented Lagrange Multiplier(ALM)method and the Alternating Direction Multipliers(ADM)algorithm are studied.At the same time,by studying the sparse representation theory,sparse coding and dictionary learning method based on the image,and analyzing the relevant model and solving algorithm,the model and algorithm of morphological component analysis based on the sparse representation are further studied to realize the different components of separation.At the same time,aiming at the problem of Gaussian noise and rain streak noise removal in WMSN video images,proposed a video image denoising algorithm based on LRTV-MCA,which can effectively remove Gaussian noise and rain streak noise in WMSN video image,and the feature information in the video image is well preserved.In the process of processing,firstly,with the advantage of low-rank sparse matrix decomposition model in dealing with sparse rain streak noise,the separation of rain streaks and video image is realized.Remove the residual Gaussian noise in video images by introducing Total Variation(TV)regularization.Then,the Morphological Component Analysis(MCA)method is used to extract the characteristic information of the video image in the process of denoising in the sparse part of the decomposition.Finally,the separated video image feature information is integrated with the low-rank part to realize the video image recovery of feature information.However,under the dynamic background condition,the effect of the proposed video image denoising algorithm is not ideal.In order to solve this problem,this paper performs the inter-frame block matching on the video,and then use the obtained similar image blocks instead of the video frames and perform the column lengths to construct the matrix to be restored.Then,by means of LRTV-MCA video image denoising algorithm,each group of similar image blocks is processed;and the image is reconstructed to realize the effective removal of Gaussian noise and rain streak noise in the video images under dynamic background conditions.