First-person hyperlapse videos

Our system converts first-person videos into hyper-lapse summaries using a set of processing stages. (a) 3D camera and point cloud recovery, followed by smooth path planning; (b) 3D per-camera proxy estimation; (c) source frame selection, seam selection using a MRF, and Poisson blending. We present a method for converting first-person videos, for ex-ample, captured with a helmet camera during activities such as rock climbing or bicycling, into hyper-lapse videos, i.e., time-lapse videos with a smoothly moving camera. At high speed-up rates, simple frame sub-sampling coupled with existing video sta-bilization methods does not work, because the erratic camera shake present in first-person videos is amplified by…