Long-time video stabilization using point-feature trajectory smoothing
Yeon Geol Ryu, Hyun Chul Roh, Myung Jin Chung · 2011
This paper presents a new video stabilization algorithm for digital video cameras. The proposed technique directly estimates correction motion through point-feature trajectories, so that our method can be free from the error accumulation as time goes on. Furthermore, an adaptive RANSAC is proposed to estimate a robust correction motion. Experimental results show that the proposed algorithm can robustly stabilize very long videos in the end.