Latency and drift are problems with tracking things. Roll and pitch are not subject to drift but yaw is. A handheld camera can move a lot between frames at 30fps. A combination of tracking data and frame tracking of the actual video might work.
Most 360 rigs are built of CMOS cameras and most CMOS cameras have a lot of rolling shutter distortion. Despite software like Warp Stabilizer in After Effects it is difficult to completely remove rolling shutter distortions in post processing.
With tracking hardware tracking accuracy and speed are inversely related. So if the system has a lot of "settle time" the accuracy can be very high.
There is a lot to be said for stabilizing the camera rig itself. The new gimbal rigs have great roll and pitch stabilizing accuracy. Whether they are compatible with the ergonomics of holding a 360 rig, over your head say, is another question. The new gimbal rigs need to in a state of very accurate natural balance before the power is turned on and it is hard to see how they will not obstruct a 360 view in multiple directions.
--- In PanoToolsNG@yahoogroups.com, Bruno Postle <bruno@...> wrote:
> On Wed 15-May-2013 at 13:07 +0200, Roberto Mancuso wrote:
> > very interesting suggetion, but you could need a device able to
> > record the inclination data....gopro doesn¹t have a similar
> > device...does it?
> Such logging hardware exists, though I have no idea if it has ever
> been connected to a 360° video rig... I was just noticing that
> stabilising these rigs mechanically has lots of really obvious
> drawbacks compared to post-processing.
> >On Tue 14-May-2013 at 10:27 +0200, Giorgio Marchetto wrote:
> >> What's the new challenges (critical points) for us with 360 video?
> >> I would like to share with you about this. What do you think
> >> about? Here some points, add yours...
> >> - video stabilization: the new promising brushless technology, a
> >> good traditional steadycam, or a gyro stabilizer?
> >It shouldn't be necessary to stabilise the entire rig with this
> >heavy equipment, a better solution would be to log the inclination
> >during recording and use this information to level the video during