SBMI-LTD: stationary background model initialization based on low-rank tensor decomposition

Abstract

Initialization of background model also known as foreground-free image against outliers or noise is a very important task for various computer vision applications. Tensor deomposition using Higher Order Robust Principal Component Analysis has been shown to be a very efficient framework for exact recovery of low-rank (corresponds to the background model) component. Recent study shows that tensor decomposition based on online optimization into low- rank and sparse component addressed the limitations of memory and computational issues as compared to the earlier approaches. However, it is based on the iterative optimization of nuclear norm which is not always robust when the large entries of an input observation tensor are corrupted against outliers. Therefore, the task of background modeling shows a weak performance in the presence of an increasing number of outliers. To address this issue, this paper presents an extension of an online tensor decomposition into low-rank and sparse components using a maximum norm constraint. Since, maximum norm regularizer is more robust than nuclear norm against large number of outliers, therefore the proposed extended tensor based decomposition framework with maximum norm provides an accurate estimation of background scene. Experimental evaluations on synthetic data as well as real dataset such as Scene Background Modeling Initialization (SBMI) show encouraging performance for the task of background modeling as compared to the state of the art approaches.