Flow-edge Guided Video Completion



Results on stationary screen-space mask completion


 
Input