Hi--

>Read the article on your site. Still not convinced. This is not my
>experience at all. Also, do this experiment take a globe, paint it
>white, project a fisheye slide on it. You will have focusing problems,
>of course, but the interesting part the apparently curved lines at the
>edges of the image, as you move around the globe-screen, are perfectly
>straight on the curved surface (actually, they are *close* to straight,
>you won't get the perfect focal for the globe, + you are starting with a
>flat slide, not actual objects). Now, if you transfer this experiment to
>how your eye functions, and I'm not saying the human eye is perfect in
>that or any other measure, it makes sense that the barrel distorsion
>caused by the lens is corrected by matching more or less the curvature
>of the retina.

Uh-uh...! ;-)

Point 1 -
The lens in the eye actually has relatively low distortion (uh, it
has pretty much uncurved, "rectangular" perspective...;-), being a
single-element design. Consider it a pin-hole, for practical purposes.
It projects onto the inside of a hemisphere, but from a point a bit
ahead of the edge of the hemisphere (indicating that the eye lens may,
indeed, have a bit of barrel distortion, otherwise a 180 degree
projection would not be possible with this geometry) - so your point
may have a bit of validity as I think of it (though, as I said before,
the projection is unlikely to be fully spherical because of the eye
geometry, though strong barrel distortion in the image must exist).

Point 2 -
In theory, for many reasons, a 180 degree image without considerable
curvature of off-center straight lines is impossible (things like
the requirement for infinite "film plane", infinitesimal FL, EXTREME
illumination fall-off, etc. preclude near "distortionless" views near
180 degrees, and we actually see wider than 180 degrees (which
essentially requires curvature in the image perspective...).

Point 3 -
The eye structure and seeing process tend to conceal the nature of the
perspective we see in, but as I point out in the article, there are
indicators of our seeing reality (the awkwardness of super-wide rectangular
perspective, though it is far less wide than our vision, for instance...).

Point 4 -
As I point out in the article, it is actually fairly easy, with practice,
to see the curvature... (once you see it, the argument is at an end...;-)
Maybe the easiest way to see it is to go to the beach and look up while
looking toward the water horizon. You can't mistake the curvature of the
horizon line for anything but a curve! ;-)

>Sorry, I forgot the other thing. If I put my nose to a 20 inch blow-up
>of a fisheye shot, I am NOT presented with the eyes' perspective.

I was refering to your example of the curved fisheye photo, which would be reconstituted into a "non-distorted" image, to be reinterpreted by the eye
into a semi-fisheye view...;-) I was pointing out that "normalizing" a
fisheye image (by curving it) was not necessary, since viewing this kind
of photo this way just returned us to the perspective of the actual
scene - either of which is then seen about the same way by the eye/brain.

>If
>there is an archway, for instance, in the shot, it looks round in the
>picture, but if I were there it wouldn't. Think of it even if I
>accepted your idea that the eyes perceive objects as barrel-distorted at
>the edges, which I don't, then this distortion would be ADDED to the
>fisheye picture, wouldn't it?

Yes - but I don't see the problem with this...;-)

>You can't have your cake and eat it too.
>The only way that uncorrected fisheye recreates reality (more or less)
>is if you print it inside a sphere, and then look at it from the
>mid-point in the sphere. And as I explained, if you then turned to the
>edges, the curved lines of the picture will appear as straight on the
>curved surface.
Stephane Leman-Langlois

Ah, but there are TWO issues here...! One is, making an image that is
seen the same way as "reality". The other is, making an image that is
approximating the way we see. These are QUITE different!!! ;-) In the
first case, we have done nothing to see how we see (reality perfectly
recreated is indistinguishable from reality, so it is equivalent...;-).
In the second, we have created a representation of how we see by making
a fisheye photograph. This is not the same process at all...
I suggest a trip to the beach...;-)