Robust optic flow algorithm based on image decomposition

Variational methods are among the best performing techniques for computing the optical flow, such methods determine the optical flow field as the minimizer of a suitable energy functional, which consists of two terms: a data term and a regularization term. In this paper, a new variational optical flow approach is proposed which design model by introducing the theory of image decomposition. First, we design the data term by texture image and structure image which could be got from decomposition model. Such method imposes invariants on image structure features instead of brightness features to improve the accuracy under varying illumination. Second, the regularization term which induces a smoothness constraint has been designed by flow- driven in order to preserve the discontinuities of flow field more accurately. Finally, combining the two terms, an advanced optical flow model has been developed. And we also present an efficient numerical scheme for solving the model. Experiments with both synthetic and real-world data show the favourable performance and the illumination robustness of our approach.