In article <4uj0va$fk3@Usenet.Logical.NET>, golem@Capital.Net says...
>Bob Neuman (d_ruether@hotmail.com) wrote:
>: In article <4ugf9q$pgd@Usenet.Logical.NET>, golem@Capital.Net says...
>
>: > I believe you are actually discussing the effects
>: > of oblique projection, whereby the magnification
>: > increases as you move off-axis. True teles have
>: > more oblique projection toward the edges simply
>: > because they are close to the film than regular
>: > long focus lenses. Inverted teles, or retrofocus
>: > wideangles, being farther from the film than the
>: > merely widefield [but non-retro] wide lens will
>: > exhibit less oblique projection effects. You can
>: > reasonably think of it as the projected image
>: > keystoning toward the corners. This is usually
>: > noticed as the egghead effect.

>: Are you saying that the rendered perspectives of, say,
>: a 15mm non-retro type and a retro-focus type super-wide
>: lens (or a 200mm tele and a 200mm long-focus type lens)
>: on the same film format would be different? I think that
>: the image of any lens of any design of a given focal-length,
>: focus, perspective, and film size is the same regardless of
>: incident angle on the film (assuming flat film....).
>: Hope This Helps

> Why assume flat film ? Actually, if the film were
> bowl-shaped, what you think about lens design, FL,
> and perspective would be all the more true.

That's true, but it would have moved the discussion away from what
you had been talking about. Technically, I could have said in the
above, "...and the same film size and shape, whatever it is...."
(Um, this could get long, and I think some of what you say below
[and above] is based on experience [and a misunderstanding], with
agreement at the end, but here goes....)

> With flat film, what you think is true is quite
> true on-axis. This means that if two utilty poles
> are pictured and one is 2X as distant as the other,
> on will look twice as tall as the other on film,
> regardless of distance, for same height poles. The
> lenses can be long or short, the images large or
> small. but the 2X dif in image height of the two
> poles will be constant. This defines a constant
> perspective, as you note.
>
> The above scenario works on-axis, even near-axis.
> If we view a row of poles receding from us on-axis
> all is in compliance with expectations. If we are
> at about 45 degr to the row of poles so that one is
> images near the center of the frame and on near the
> edge, a Super Angulon or similar non retrofocus WA
> will render the pole near the format edge larger
> than it will render it if we pan the camera to put
> the same pole thru the center. An SLR WA [retro]
> will reduce this effect.

I don't think the difference is true, IF you equalize the linear
distortion in the two lenses, as I did in the conditions I set
in the paragraph above, "...focus, *perspective*, and film size..."
(perspective includes linear characteristics, since it really
defines them - IOW, a truly "rectangular" lens type renders all
subject straight lines straight in the image (assuming flat
film....;-). You may have observed the differences in perspective
effects that you described above when using a super-wide
retro-focus SLR lens, but I would attribute it to the considerable
barrel distortion present in that particular SLR lens (and maybe
to a little pincushion present in the Super Angulon near the edge
of coverage), but not to the difference in lens design (and
relative spacing from the film). If you compared the image effects
in, say, a Nikkor 15mm f5.6 (with relatively little linear
distortion) with a 90mm f8 Nikkor (or 15mm Zeiss Hologon) over
the same angle of view, I think you would find little difference
in perspective effects, and what little difference remained
could be accounted for by the differences in linear distortion.

> We can be at 45 degr to the line of poles, 50 ft
> from the pole imaged thru center and 25 feet from
> another pole images near the format edge. The 2X
> image size expectation will fail with both lenses,
> but fail more greatly with the Super Angulon than
> the SLR lens. Witha normal lens the failure is
> reduced because we can['t] get as far off-axis. With
> long lenses, the off-axis angle is so minor that
> the dif between true tele and long focus is mostly
> academic.

Here we need to be careful of planes - if the center pole is in
a plane perpendicular to our line of sight and 50 feet away, and
the second pole is in a plane also perpendicular to our line of
sight (and therefore parallel with the first) and that plane is
25 feet away, the 2X observation would continue to work (and the
second pole nearer us would be about 35.36' away from us), and
the perspective would still be the same for both lens types,
if they are free of perspective type-altering linear distortion.
As you state it, the 2X condition will fail (since the plane
distance relationship has been disturbed [you have rotated the
pole distance around us instead of sliding it along the plane,
giving us a different perspective type {"cylindrical", as in a
swing-lens curved-film panorama camera - in which an object
does not change size with angle from center as long as the
distance from center is kept constant}]).

> "Oddly" enough, this effect won't plague a fisheye.
> If the near pole is near the format edge and the
> doubley distant pole is thru center, the near pole
> can easily be rendered smaller than the doubley
> distant pole, or same-size, or other ratios, all
> by a tiny re-aiming of the camera while standing
> in the same place. Fisheyes have variable
> perspective across the format.

No, technically, most have constant "spherical" perspective, and
effects seen anywhere in the field follow the rules of spherical
("fisheye") perspective. An object anywhere in the field at a
constant distances from us will maintain constant size, very
unlike what happens in rectangular perspective in any direction,
and unlike what happens in cylindrical perspective when moving an
object vertically (assuming horizontal rotation).

> A Widelux or other sweep-lens panoramic camera will
> follow formula for the image heights of the poles
> regardless of distance or whether we aim down the
> row of poles nearly head on or allow them to span
> the whole format edge-to-edge. This is because all
> imaging is on-axis when you have a curved film
> swept by a moving lens axis that is the radius of
> the curve of the film. Horizontally, there is no
> off-axis imaging in a Widelux. Vertically, there
> is little dimension to the format anyway.
>
> The lack of error in the curved film camera
> demonstrates the off-axis error that occurs in a
> flat film camera. aim the Widelux at any single
> pole and you can place that pole anywhere across
> the format and it will remain the same image height.
> Try the same with a 65mm on a 4x5 and the pole will
> grow taller as you aim to place it away from center.
> This is a natural result of changing the angle of
> the film plane relative to the lens axis. We can
> use this toi advantage if we combine it with other
> movement on a view camera, but the principle of
> oblique projection is the same in both cases. On
> the Widelux, when you swing the the "film plane"
> nothing happens because there is no single plane in
> which the film lies. It has a curved film plane.

There is no "error" in the flat-film camera with the
rectangular-perspective lens - the unexpected off-axis effects
are normal to the type (and are identical for retro-focus,
standard, and telephoto lens design types), but their strangeness
to us does argue for our seeing in a different perspective type.
It may be easier to "see" all of this if we forget about two poles
at different distances from each other, and move one pole through
the field of coverage, using the rules of the different
perspective types.

> Flat film cameras with "rectilinear" lenses lack
> the abilty to render true perspective. If you
> must use flat film, the more barrel "distortion"
> allowed, the more accurate the perspective, right
> out to full fisheye. Curved film can render even
> more accurate perspective, but the simple curve of
> a pano camera corrects only horizontal error [when
> in "landscape mode"]. A bowl shaped "film plane"
> combined with pinhole optics would allow completely
> true perspective rendering, assuming also a bowl
> shaped print from the bowl shaped neg !! But a
> flat print from a flat neg just preserves the
> error already extant in the flat neg.
>
> David Rosen golem@capital.net

If by "true" perspective you mean, "what perspective most closely
resembles what we use when seeing", I would hartily agree. (And
there is no need to curve the film to achieve spherical perspective
rendering on flat film [unlike cylindrical perspective] - the
fisheye lens does it nicely.) Your bowl-and-pinhole camera closely
resembles the eye structure - which is the reason, I think, for the
"rightness" of most fisheye effects as opposed to rectangular ones,
when the angle of view gets wide. (With the unfamiliar effects
being fairly easy to explain away, as I hope I successfully did
in the original post, "On seeing and Perspective". ;-)

BTW, in all of this, I trust that you (and the other poster, DG)
do not take this in a competitive spirit - I do not intend my
posts that way. This is an interesting subject to me, and I
respond, since I may have something to offer.....
Hope This Helps d_ruether@hotmail.com