Benefits of temporal oversampling in optical flow estimation

Recently it has been shown that optical flow estimation (OFE) accuracy can benefit from temporal oversampling especially for large displacements between frames. In this paper, we show that temporal oversampling also benefits OFE in the case of complex scenes with small displacements but high spatial bandwidth. Using synthetic test sequences and a high-speed real video sequence, it is shown that temporal oversampling improves the performance of OFE by reducing motion aliasing not only for areas with large displacements but also for areas with small displacements but with high spatial frequencies. We also demonstrate that the minimum frame rate necessary to achieve good OFE performance for the tested sequences is largely determined by the minimum frame rate necessary to prevent motion aliasing.