I agree with the others, lets keep the discussion here.
I have asked to give a talk on stereo stitching at the IVRPA meeting in Prague
On 2015-03-16 2:34 AM, Roger Williams wrote:
> Aha! BIG AHA!!
> I do know about horizontal and vertical line control points but I also knew from experience that the horizontal line normally has to be the true horizon, as this is the only line that is straight in an equirectangular image. But I now see that they are perfectly capable of identifying points that DO lie along straight lines corresponding to the linear separation of the cameras. Nice miscommunication!
> Now I think I've got it. I imagined you firing away with twinned cameras. Now I know you are like me, faking it with a single camera, I am greatly encouraged. I have been using (or rather trying to use) Wim's technique with a single camera masking out all but a wedge on the LH and RH sides of the fisheye images. Still not there yet. It requires rotation in a true horizontal plane as near perfectly as possible. I am not nearly there with my shaky camera-perched on-monopod-on-top-of-rotator-on-a-mini-tripod.
I think a sturdy tripod is a must.
> I do have a slide bar that will allow me to move one camera the 6.5 or 7cm needed.
My example of the back deck only had about 25mm separation. I hope to experiment
with different separations.
> I am tempted to use a larger, firmer tripod than my normal setup to achieve more reproducible imaging, although I'm too old to want to lug lots of heavy stuff around. Wife does not take kindly to being my Sherpa. I am thinking of using P2VR to limit the angles of view so that no-one gets to see the messes I am leaving at the zenith. There really is very little a most zeniths, and my feeling is that the "wow" of the stereo effect should be enough to please people even if they can't look straight down... How do you feel about that? In the coming era of immersive VR it may be important to allow people to look all around, I suppose. But maybe by then Joost will have developed 3D image field techniques that can solve the parallax problems of 3D stitching? (hint, hint!)
I think you want to still look down. Having edges and boundaries takes away from
the immersion. When looking on a HMD I had to extend the tilt so I could look
past the zenith and nadir. That can be disorienting on the monitor with a mouse
but with a HMD your head know where it is looking and having a limit is very
If I had a wider fisheye lens them I would extend the stereo further. But I
limit it to what my 10.5 on cropped sensor can see. I could shoot two rows but
that is starting to get complicated. Anyone serious about doing stereo I think
should use two cameras to freeze motion.
> Roger W.
> PS Apologies to people for this extensive exchange on a minority interest. It's not exactly Off Topic, but maybe we should take it to Personal Mail?