Comment by hdersch
7 hours ago
I wonder how this compares to purely vision-based systems which use nothing but the images themselves for stabilization. Here are some quite old results of stabilization using image-based 3d-reconstruction of the scene which I wrote more than 10years ago, compared with other stabilization programs of that time (Deshaker, Adobe After Effects, Youtube). With todays improved hardware and progress in 3d-algorithms you may not need any additional gyroscopic data.
I think gyroscopic data still can have the edge if it has higher sampling rate than the video: then it could be used for removing blur from individual frames.
I also expect purely gyroscopic approach to be much lighter compute-wise.