Hyper-Lapse From Multiple Spatially-Overlapping Videos
This publication appears in: IEEE Transactions on Image Processing
Authors: M. Wang, J. Liang, S. Zhang, S. Lu, A. Shamir and S. Hu
Number of Pages: 13
Publication Year: 2017
Hyper-lapse video at high speed-up rate is an efficient way to illustrate long videos such as a human activity in first-person view. Existing hyper-lapse video creation methods produce fast-forward video effect from only one video source. In this work, we present a novel hyper-lapse video creation approach from multiple spatially-overlapping videos. We assume the videos share a common view, location or subject, and find transition points where jumps from one video to another can occur. We represent the collection of videos using the hyper-lapse transition graph where edges between nodes represent possible hyper-lapse frame transitions. To create a hyper-lapse video, we run a shortest path search algorithm on the digraph that optimizes the frame sampling and assembly simultaneously. Finally, we render hyper-lapse results using video stabilization and appearance smoothing techniques on the selected frames. With the proposed technique, one can synthesize novel virtual hyper-lapse routes which may not exist originally. We show various application results on both indoor and outdoor video collections with static scenes, moving objects as well as crowds.