Generating spherical hyperlapse videos via recursive intelligent sampling for StratoJump

Author

Higuchi, Keita and Kasahara, Shunichi and Nitta, Kei and Yanase, Yohei

Abstract

StratoJump is an installation that allows participants to explore the experience of high altitude jumping toward the edge of space. In this peper, we presents an vision-based algorithm for generating hyperlapse of spherical videos captured in high altitude space. Unlike previous work, we consider high altitude videos that cannot perfectly be stabilized by 3D reconstruction-based algorithms. We proposed recursive intelligent sampling to simultaneously stabilize and shorten spherical videos. Our preliminary results show that the proposed method can reduce cumulative errors of stabilization compared to a frame by frame method in reasonable time.