Spatially and Temporally Optimized Video Stabilization

Abstract

Abstract-Properly handling parallax is important for video stabilization. Existing methods that achieve the aim require either 3D reconstruction or long feature trajectories to enforce the subspace or epipolar geometry constraints. In this paper, we present a robust and efficient technique that works on general videos. It achieves high-quality camera motion on videos where 3D reconstruction is difficult or long feature trajectories are not available. We represent each trajectory as a Bé zier curve and maintain the spatial relations between trajectories by preserving the original offsets of neighboring curves. Our technique formulates stabilization as a spatialtemporal optimization problem that finds smooth feature trajectories and avoids visual distortion. The Bé zier representation enables strong smoothness of each feature trajectory and reduces the number of variables in the optimization problem. We also stabilize videos in a streaming fashion to achieve scalability. The experiments show that our technique achieves high-quality camera motion on a variety of challenging videos that are difficult for existing methods.

Video stabilization based on low‐rank constraint and trajectory optimization

“…However, their method performs poorly when dealing with videos that contain large foreground objects and parallax scenes. Liu et al [23] employed subspace constraints to smooth the path trajectories and utilized contentpreserving warps [24] to generate stabilized frames, thereby improving the robustness of video stabilization for parallax scenes. Zhao et al [25] proposed an adaptive grid partitioning method based on feature trajectories, while Goldstein et al [26] modelled stabilized camera motion using a fundamental matrix and employed epipolar geometry to determine the locations of smoothed trajectories.…”

“…Liu et al. [23] employed subspace constraints to smooth the path trajectories and utilized content‐preserving warps [24] to generate stabilized frames, thereby improving the robustness of video stabilization for parallax scenes. Zhao et al.…”

Robust and computationally efficient online image stabilisation framework based on adaptive dual motion vector integration

“…Digital image stabilisation methods can be broadly classified into three categories: 3D, 2.5D, and 2D methods, based on the motion model employed. The 3D [28][29][30] and 2.5D [11,[31][32][33] methods achieve notable performance particularly in scenes with large depth variations. However, the aforementioned methods are unhealthy for online image stabilisation mainly due to two reasons: (i) their computational intense nature, and (ii) requirement of future frames.…”

Sequential and Automatic Image-Sequence Registration of Road Areas Monitored from a Hovering Helicopter

“…Liu et al [ 16 ] determined 2D feature trajectories, which were consecutively smoothed to produce a stabilized image sequence of a quality comparable to the results of the 3D method of [ 14 ]. Wang et al [ 17 ] represented trajectories by Bezier curves. For these methods, long trajectories are required.…”