Visual Tracking via Bilateral 2DPCA and Robust Coding

The success of sparse representation, in face recognition and visual tracking, has attracted much attention in computer vision in spite of its computational complexity. However, these sparse representation-based methods often assume that the coding residual follows either Gaussian or Laplacian distribution, which may not be precise enough to describe the coding residuals in real tracking situations. With the aim of dealing such coding residuals, in this paper, a novel generative tracker is proposed in a Bayesian inference framework by exploiting both the bilateral 2D principle component analysis (B2DPCA) and robust coding. As the coding residual is two-dimensional, the weighted residual minimization is extended and is introduced into B2DPCA reconstruction without considering the much complex l(1)-regularization. Further, it is to proposed use the weights obtained during the process of residual minimization for generating an occlusion matrix, which is used to enhance the tracker updates. The proposed method is evaluated on the challenging image sequences available in the literature, and it is demonstrated that the proposed method performs favorably when compared with the several state-of-the-art methods.