Background Extraction From Video Sequences Via Motion-Assisted Matrix Reconstruction

In recent years, high-definition video technology rapid development and widespreadapplication. So the digital video compression, processing, transmission, storage and otheraspects proposed the new challenge. Digital video capture scenes including the foregroundand background, the foreground and background separation can reduce many videoanalysis of the difficulty and complexity, thus becoming an important research direction invideo processing. This paper makes use of the latest matrix recovery and filling theory,combining the motion information of video sequences, background extraction method, themain research results are as follows:1) This paper first introduces the optical flow based method for motion estimation.Then, analysis the effect of video compression on the motion information to estimate theinfluence of the optical flow field, obtained accuracy and coding parameters betweenassociation, based on two kinds of new coding H.264/AVC and HEVC.2) Presents a method based on motion information and matrix recovery videobackground extraction method. The method of optical flow field of two values, isestablished by motion information weighted observations of the matrix, to establish arobust principal component analysis model. Finally, using convex optimization method tosolve it, restored background image. The simulation results show that, the proposedmethod in various types of video sequence in the stable and accurate extraction of thebackground, and is better than no motion-assisted robust principal component analysismethod.