SteadyFlow: Spatially Smooth Optical Flow for Video Stabilization

We propose a novel motion model, SteadyFlow, to repre-sent the motion between neighboring video frames for sta-bilization. A SteadyFlow is a specific optical flow by en-forcing strong spatial coherence, such that smoothing fea-ture trajectories can be replaced by smoothing pixel pro-files, which are motion vectors collected at the same pixel location in the SteadyFlow over time. In this way, we can avoid brittle feature tracking in a video stabilization sys-tem. Besides, SteadyFlow is a more general 2D motion model which can deal with spatially-variant motion. We initialize the SteadyFlow by optical flow and then discard discontinuous motions by a spatial-temporal analysis and fill in missing regions by motion completion. Our experi-ments demonstrate the effectiveness of our stabilization on real-world challenging videos. 1