Research Of Video Repair Algorithm Based On Low Rank Representation

In recent years,digital video has been widely used in life,scientific research and other aspects.However,due to the low quality of video files or the existence of noise and other serious impact on its normal viewing and practical research,the research on video repair has great practical significance and application value.Low-rank matrix restoration algorithm(LRMR)is widely used in video restoration as a dimensionality reduction algorithm.Its main idea is to represent the observed data matrix as low-rank part and sparse part,so as to retain the low-rank part close to the original data for analysis and application,thus achieving good results.However,the existing video repair algorithms based on low-rank representation still have problems such as high computational cost,sensitivity to outliers,and algorithm instability.In view of this situation,this paper proposes a new improved video repair algorithm based on the research of the existing video repair algorithms.The improved video repair algorithm is used to realize multi-frame joint automatic repair without reference,the repair problem of each frame is transformed into a low-rank matrix recovery problem.By factoring the robust PCA and adding strong convex regularization term to perform singular value correction,the speed and accuracy of separating the sparse error from the original content were improved.Experimental results show that the proposed algorithm can achieve multi-frame joint on reference and unsupervised video restoration without forecasting the area to be repaired and the noise model.The main work of this paper includes:Based on the detailed study of low-rank matrix restoration(LRMR)and low-rank representation(LRR),bilinear factorization of RPCA model and strong convex regularization terms are applied to propose the proposed algorithm,which is a stable Frobenius/Nuclear mixed norm repair algorithm based on low-rank representation.To take advantage of low rank part between consecutive frames in video significantly correlation between and defects,noise randomness has nothing to do with sex,for video frame joint batch implement automatic recovery,the resulting corresponding low rank matrix and sparse matrix,at the same time for the algorithm is suitable for various kinds of video processing and harmonic parameters improvement,after using the harmonic parameters model for repair.Finally,the processed video frame matrix is reconstructed to get the reconstructed matrix,which is outputted in the form of video,so as to realize the approximate restoration of video.Finally,through Matlab simulation,the proposed algorithm is applied to standard video library and real public data set Pits to carry out experiments.The experimental analysis shows that the proposed method can obtain higher quality repair results,even in the case of large noise ratio.The experiment of standard database proves that the method presented in this paper has strong generality and good performance in various types of video restoration,such as remote view,close view,people,landscape,traffic,etc.Experiments with real data verify the basic generality of the proposed method,which can be used as the basic theoretical support for some practical application problems,such as background modeling,foreground segmentation,etc.In the last two experiments,the results of data analysis and comparison with other algorithms show that the proposed algorithm is usually superior to the latest methods.