The conventional optical flow has a fundamental limitation in handling motion details and image registration. In this paper, we propose a Zernike moments descriptor matching based symmetric optical flow estimation for high-quality image registration and motion estimation, which is an
