Method of Optimal Caemra Path Estimation Based on Optical Flow for Video Stabilization

Video captured by a moving camera system often suffer from annoying jitters because of undesired motion. This paper presents an optimal camera path estimation method of moving camera for stabilized video generation. The proposed video stabilization algorithm consists of three steps: i) estimation of optical flow between adjacent frame, ii) extraction of original camera path based on optical flow, and iii) optimal camera path estimation and frame warping. Experimental result show that the proposed video stabilization algorithm can significantly remove the undesired motion. The proposed method can be applied to surveillance system, driving assistant system, and various computer vision systems.