Recovering low-rank and sparse components of matrices for object detection

It is shown that object detection can be addressed in the authors' unified framework, where the observed video matrix is decomposed into the low-rank matrix and the sparse matrix. The recovering problem can be solved by the proposed variant of the Douglas-Rachford splitting method, which accomplishes recovery by exploiting the separable structure property of the model. The effectiveness of the proposed object detection scheme is illustrated on two data: simulated data and real sequences applications. The numerical experiments verify that the proposed algorithm has attractive robustness and high accuracy for illumination variation and dynamic texture.