Background modeling from video sequences via online motion-aware RPCA

Background modeling of video frame sequences is a prerequisite for computer vision applications. Robust principal component analysis(RPCA), which aims to recover low rank matrix in applications data mining and machine learning, has shown improved background performance. Unfortunately, The traditional RPCA method considers the batch recovery all samples, leads higher storage cost. This paper proposes novel online motion-aware algorithm, named OM-RPCAT, adopt truncated nuclear norm regularization as an approximation constraint. And then, Two methods are employed obtain motion estimation matrix, optical flow selection, merged into items separate foreground background. Finally, efficient alternating optimization algorithm designed manner. Experimental evaluations challenging demonstrate promising results over state-of-the-art application.