Output details

Incremental Tensor Subspace Learning and Its Applications to Foreground Segmentation and Tracking

<23> This paper is one of several resulting from a long standing collaboration with Institute of Automation, Beijing. It describes the first online incremental tensor subspace learning algorithm for foreground segmentation and tracking. The incremental learning ensures that the algorithm can be applied in real time as each new image in the sequence arrives. This work inspired at least three subsequent papers: “Tracking by Third-Order Tensor Representation,” IEEE Trans. on Systems, Man, Cybernetics, 2011; “Incremental Learning of Weighted Tensor Subspace for Visual Tracking,” IEEE Conference on Systems, Man, Cybernetics, 2009; "Discriminant Tracking Using Tensor Representation with Semi-supervised Improvement", ICCV 2013.