Bundled camera paths for video stabilization

Abstract

a) a single global path (b) our bundled paths Figure 1: Comparison between traditional 2D stabilization (a single global camera path) and our bundled camera paths stabilization. We plot the camera trajectories (visualized by the y-axis translation over time) and show the original path (red) and the smoothed path (blue) for both methods. Our bundled paths rely on a 2D mesh-based motion representation, and are smoothed in space-time. AbstractWe present a novel video stabilization method which models camera motion with a bundle of (multiple) camera paths. The proposed model is based on a mesh-based, spatially-variant motion representation and an adaptive, space-time path optimization. Our motion representation allows us to fundamentally handle parallax and rolling shutter effects while it does not require long feature trajectories or sparse 3D reconstruction. We introduce the 'as-similaras-possible' idea to make motion estimation more robust. Our space-time path smoothing adaptively adjusts smoothness strength by considering discontinuities, cropping size and geometrical distortion in a unified optimization framework. The evaluation on a large variety of consumer videos demonstrates the merits of our method.

Selectively De-animating and Stabilizing Videos

“…The third system, User-assisted Video Stabilization (Chapter 5), seeks to incorporate content information and user goals into video stabilization. We use the state-ofthe-art Bundled Paths [77] automatic video stabilization as a baseline and build upon their work to include the user in the loop.…”

“…Video stabilization techniques have also been developed for light fields [99] and for videos with depth information [100]. Bundles of homographies guiding cells in a grid mesh achieve state-of-the-art results by optimizing the path of each grid cell while enforcing spatial coherence [77].…”

“…Our work directly addresses these two issues by providing two tools for the user to customize the stabilization result. Our algorithm starts with an initial baseline automatic video stabilization result [77] which the user can further improve using our tools. First, the user can improve track selection to ensure that only background tracks are selected.…”

“…Bundled paths [77] is a state-of-the-art video stabilization algorithm which performs well with challenging videos. The algorithm divides each frame into a 16 by 16 grid and estimates camera paths throughout time for each cell.…”

“…These approaches cannot de-animate large-scale motions of specific objects while leaving the finer-scale motions intact. Video textures [97,65,7,77] are a well-known approach for seamlessly looping stochastic motions. Like cinemagraphs, a video texture lies between photography and video, but a video texture does not suppress distracting motions within the frame.…”