PDA

View Full Version : Vertical parallax and calibration.

Ion
09-21-2008, 07:16 AM
Hi everyone (and John; I&#039;ll be needing you again),

* I have been working on the calibration of my D300 + Nikkor 10.5 on my NN5L.
Things work out oke but I&#039;m still puzzled with the no-parallax-point calibration for the zenith shot.

I use a -15 angle for smaller nadir footprint. I am wondering what the &#039;optimal&#039; zenith angle would be. I have been shooting +60 / +90 and +75 degrees.
Actuall the +60 gave me best (at least easiest) results when stitching outdoor scenes with clouds or air cause you have a reference &#039;on the ground&#039; for control points.

I have been thinking (correct me if I&#039;m wrong) if it would be more correct to shoot at a +75 degrees angle. -15 + 90 = +75 degrees (similar angle as if you would use 0 degrees and a 90 degrees zenith shot).
Or would 60 degrees increments work better (same angle used for the vertical shots), for example +45 degrees (but here the covered area will be too small I think) or +105 (so -15 actually) or is this mathematical mumbojumbo? :-)

For vertical parallax should the same method be used to find the NNP and rotate the camera around the horizontal axis with +-45degrees (or +-30 degrees when using a 60 degrees angle for the zenith shot)

* Secondly... Hans Nyberg can confuse you sometimes too you know haha :-)
http://www.panoramas.dk/panorama/fisheye-NPP/parallax.html

I suppose this is applicable too the 10.5fish as well? So it&#039;s not only the angle used for the photo&#039;s (the amount of shots to cover the 360degrees) but also the place on the photo itself where you want to avoid parallax, that are important to find the "NNP". What influence does Hans&#039;s theory have on your &#039;classic&#039; (with the tape on the window) method for finding the NNP, John?

The more I learn and think I understand about pano&#039;s, the more I discover that has to be learned and done.

Best regards, Wim.

John Houghton
09-22-2008, 03:59 AM
Given the way that the entrance pupil is smeared over a surface rather than being a simple point, it is impossible to eliminate parallax perfectly. Mostly, the places where parallax effects show are on the actual seams, and the light rays from subject features at the seams will not be entering the lens at a constant angle to the optical axis. The most favorable shooting scenario is a Sigma 8mm, say, with 4 shots around + Z + N. The horizontal images will join at yaw +/- 45 degrees. So too will the zenith and nadir join with the horizontal shots at +/- 45 degrees of pitch. Therefore setting the no-parallax position to suit a joining angle of 45 degrees will work reasonably well at all the seams - both horizontal and vertical.

When the camera is tilted up or down at varying angles, the seams are going to occur ar various angles and there will be no NNP that is ideal for all of them. Compromises have to be made. You can adjust the no-parallax position to suit those seams where parallax effects are likely to be more noticeable. This might be the floor, which is both near and at a big angle of pitch in a horizontal fisheye shot. At the horizontal level, things might be fairly distant, and therefore less affected by parallax. OTOH a nadir shot would cover any stitching errors down there, so maybe that&#039;s a way to overcome potential problems there.

The best plan is to set up the pano head as best you can using any of the esatblished methods and than fine tune the position in the light of experience by shooting panoramas for real. Stitching software is very forgiving these days, and even if there are occasional stitching errors, they can usually be easily and quickly corrected.

John

Ion
09-24-2008, 04:53 AM
Hey John and the rest of the panofreaks.

Thanks again for the help.
Seriously you should write a (e)book or so John, I&#039;d be happy to spend some money on this kind of valuable information.

I&#039;m trying out -10degrees now and 90degrees zenith shot. I think this will &#039;even&#039; out parallax errors a bit overall.
See y&#039;all.
Wim.