(Page 1 of 2 pages for this article 1 2 >)
Tuesday, August 05, 2008
Make the Entrance Pupil Your Friend
Eliminate parallax errors on pans and tilts (for fun and profit)
I’m helping a friend build a three RED camera panoramic camera system that will shoot a 180 image, combining three 60 degree images in post. One of the more important things to consider when doing this kind of work is finding, and aligning, the “pivot point” of each lens to eliminate parallax errors that can occur during panning and tilting.
Some of you may be asking yourselves, “What the hell is this guy talking about?” For those of you who think such things, let me explain:
Every lens has a pivot point around which you can rotate the lens and not see any parallax errors during panning or tilting. Commonly, and erroneously, called the nodal point (this is what I was taught to call it, but apparently that is now completely wrong), this point is defined by the location of the “entrance pupil,” or the lens’s peephole to the world. For a lens where the aperture iris is in front of the body of the lens, the location of this point is at the aperture hole itself. For a lens where the aperture falls behind the body of the lens, the entrance pupil could be anywhere--inside the lens or, occasionally, in front of it. In a prime lens this location will be fixed; on a zoom lens it can move around with focal length changes.
Imagine a simple setup where we are shooting two C-stands, lined up in a row so that one blocks the other:
The camera is balanced on a fluid head (not shown). Watch what happens when we pan:
The rotation, or pivot, point of the camera/lens system is under the camera, so that the lens physically moves to another location when panned. That results in what many consider to be simply a side effect of wide angle lenses, which is that the foreground appears to move faster than the background. This movie, shot in the prep area at Chater Camera in Berkeley, demonstrates this effect using a RED ONE, an 18mm Arri Ultra Prime, and two headless C-stands: one parked about two feet from the camera, and one placed directly behind it about six feet away, both with orange tape attached:
Notice how, as the camera is panned, the rear C-stand becomes visible. That’s called parallax error, and that’s not always desirable. For example, in days of old it was common to execute special visual effects in-camera, in real time, because computers didn’t exist yet. One common trick was to paint a background, for example a castle, on a hill, on a piece of glass, leaving a clear area through which actors would be shot approaching. Most of these shots were locked off, but occasionally it was beneficial to be able to pan with the actors to help sell the “reality” of the shot. If the camera was panned on the lens’s pivot point then all the shot elements would track properly.
It’s also beneficial to pan around the pivot point in panoramic photography, as the images can be stitched together more cleanly without close objects shifting in odd ways that give away the stitching.
How to find the pivot point, and the result of doing so, on page 2…
(Page 1 of 2 pages for this article 1 2 >)
Mark Christiansen | 11/18- 05:31 PM
Art Adams | 11/16- 06:41 PM
Art Adams | 11/14- 01:44 PM
Randy Boyes | 11/14- 09:12 AM
Art Adams | 11/13- 01:23 PM
Art,
Great article as usual! Thanks man, keep them coming.
Posted by on 08/05 at 09:58 AM
good article…
so to use 3 red cameras
and not have parallax,
you have to either have them , by some miracle, inhabit the same space
or
shoot through mirrors
whats he gonna do??
i gots to know!
Posted by billS on 08/08 at 01:56 AM
|