Dense optical flow estimation with 3D structure tensor models

Abstract

Since the 3D structure tensor at each pixel can interpret the local between frames well, it can be used to estimate dense flow. According to the assumptions of brightness constancy, the optical flow estimation can be converted to the calculation the eigenvector of the structure tensor, rather than the complex calculation of linear system. Iterative coarse-to-fine refinement is used to improve the performance. Experimental results show that the proposed algorithm is robust and effective for computing the dense flow.