Re: Keystoning/Ghosting

[P3D John Bercovitz <bercov@xxxxxxxxxxx>]

Adrian Schubert asks:

> 2) why is it that in a small theatre that shows polarized 3d 
> films everyone in the room can see the effect, regardless of 
> position? It seems (from my testing) that one needs to be in a 
> specific position for a particular arrangement of the glasses 
> and projector filters...

Was your testing done only with the sheet of aluminum which you 
mentioned?  You won't get very good light distribution from an 
aluminum sheet unless its surface is really roughed up, perhaps 
by bead blasting?

> 3) Should ALL 3d photos be taken as parallel views, or is 
> keystoning (centering a specific NEAR subject in both images 
> (non-parallel views) have advantages in some cases? I have read 
> that keystoning is bad...BUT, I've seen some awesome anaglyphs 
> that had the DISTANT objects more greatly separated than the 
> near ones....which seems contrary to the parallel-view method. 
> What is going on?! I just got back a first batch of 3d 
> anaglyphs, and most have problems, probably due to the picture-
> to-picture angle I used...WHAT is the correct technique for 
> taking stereo pairs?! I can't find any clear reference to this 
> anywhere, and I'm dying to know. More specifically, what is 
> needed to have objects seem to pop out of the image plane? 
> Longer baseline, keystoning, or simply nearer objects?

Yes, all 3D pairs _should_ be taken with both pieces of film in 
the same plane but sometimes it's hard to avoid having to use a 
little "keystone" to get enough usable image area.  I would 
recommend instead of keystoning that you reduce the stereo base to 
increase shared image area.  For an excellent discussion of why 
keystoning causes warpage of 3D space, see Andrew Woods' articles 
in the bobcat anonymous ftp/web site.  If you are unfamiliar with 
the site, write me off-list.  The address is:
ftp://bobcat.etsu.edu/pub/photo/photo-3d/technical/

In any pair of images, the corresponding image points of a distant 
object should be farther apart than the corresponding image points 
of a near object.  

To have objects pop out of the window, you have merely to make 
sure that the object's image points are closer together than the 
edges of the stereo window.  The closer together, the nearer, 
regardless of whether you're talking about images or windows.

Longer baseline has the seemingly contradictory effect of 
increasing the stereo effect while reducing the scale of all three 
dimensions (equally) in the 3D image space.  There are 
illustrations in the technical section of the bobcat site which 
show how this happens.

John B


------------------------------