Re: Computer Compositing

[T3D Larry Berlin <lberlin@xxxxxxxxx>]

john bercovitz  writes:
>You're actually talking about two different things here.  Larry is talking 
>about shifting an image part which results in a flat image and Michael was 
>talking about taking slanty views of the image part, one from the right and 
>one from the left, which results in keystoning in each view and that causes
>curved hyberbolic distortion of a flat surface.

I'm having trouble visualizing the curved distortion. All the manipulation
interpretations I've experimented with show a planar rotation but not a
distortion of the plane. Keystoning introduces a vertical compression in the
result and a corresponding vertical misalignment with parts of the original.
Where it's too great, I can't match up image parts but the basic parallax
shift still seems linear. When I simulate the keystoning without any
vertical compression, using just a horizontal compression, the image is on a
plane that has been rotated from perpendicular but not distorted in any
curved way that I can see. 

To what does *slanty views* refer, something different from a
non-perpendicular view of the flat image?  

Larry Berlin


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