Re: Poll on *Really* Wide Angle Lenses

In article <sOGdnWJzd53EaWXfRVn-3A@xxxxxxxxxxx>,
Nostrobino <not@xxxxxxxxxx> wrote:
>
>"Chris Brown" <cpbrown@xxxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
>news:o29ms2-rni.ln1@xxxxxxxxxxxxxxxxxxxxxxx
>
>> You are quite simply wrong here. Objects look smaller the further away
>> they
>> get - it doesn't matter what angle you look at them at.
>
>No, BC is correct here, at least as far as flat-field rectilinear
>photographic lenses are concerned.

I specifically said, "with his own eye". The eye is not a flat-field
rectilinear photographic lens. It's not a fisheye either. Consider two
properties we might like an imaging system to have:

- Identically sized objects which are equidistant from the viewer should
have the same apparent size.

- Straight lines should look straight.

A fisheye lens has the first property, but not the second (except in the
specific case where the line goes through the centre of the image). A
rectilinear lens has the second property, but not the first (except in the
specific case where the two objects are also equidistant from the centre of
the image).

If you think like a photographic lens designer, you'll probably know that
you can't have both these properties simultaneously.

The problem is, you'd be wrong, because human vision has both these
properties simultaneously (as does the internal representation of an image
in Panotools, interestringly, and for the same reason). You point out the
reason for this below.

>The bricks (or other details on what is
>essentially a two-dimensional surface) do not change in size regardless of
>their position, as long as the lens axis is perpendicular to the wall.

We already established that this is the case *with a rectilinear
projection*, but that's a bug, not a feature. In human vision, objects that
are further away look smaller, and ultimately the reaso9n we take
photographs is so that people can look at them.

With a narrow angle of view, neither rectilinear projections not fisheye
projections look offensively wrong. The problem comes when you get to an
extreme wide-angle view. Neither represents the world in the same way we see
it. Most people seem to find the rectilinear projection less aestetically
offensive, but I suspect this may actually just be a cultural thing - in the
developed world our mathematics education from an early age makes us think
of geometry in Catresian terms - we just don't like the idea of parallel
lines converging.

Where the rectilinear projection really does start to get offensive though,
is when you use it in motion pictures (more on this below).

>> Now turn through 90 degrees, so that you are perpendicular to the scene.
>> Continue to lok straight ahead, but observe your pervious subject using
>> your
>> peripheral vision. Notice how things still look smaller the further away
>> they are.
>
>One problem is that that's extremely difficult to do. While both eyes
>working together give us about a 180-degree horizontal field of view, only
>the center portion has enough sharpness and detail to give us much clear
>information about the image.

There's the rub. Brian probably thinks his brick-wall example really is how
we see it, because most people aren't used to looking straight on at a close
object and having the perspective effect be visually obvious, but it is
there, and if you try hard you can see it.

>But I think much the larger problem is that
>it's an apples and oranges comparison: The rectilinear camera lens projects
>its image on a flat plane. The eye's lens does not.

Bingo! That's the nub of the thing. On the surface of a sphere, parralel
lines stay straigtht, and yet still meet at both horizons. In Brain's brick
example, human vision simultaneously represents the mortar lines as
straight, *and* the bricks as appearing smaller as they move further away
from the centre of the visual field.

It's not possible to do this in a photograph, not unless you project the
image onto the inside of a sphere.

>I don't believe there is
>any way that a rectilinear camera lens can reproduce on a flat surface the
>hemispherical image formed on the retina by the eye's lens. I suspect that a
>fisheye lens actually comes closer to doing that, but seems to have
>curvilinear distortion because we are looking at its image on a wrong (i.e.,
>flat) surface at a wrong distance. The eye-brain system sorts out shapes and
>sizes in three-dimensional space, in a way that I don't think any camera
>lens can do.

After a fashion. I think a better way of looking at it is to say that
fisheyes and rectilinears both chose to make a different compromise. The
main benefit of a fisheye is that you aren't limited in your field of view
in the same way - you can go to (and beyond) 180 degrees.

This seems like a good point to come back to the offensivness of rectilinear
lenses in motion pictures. Have you ever seen one of those films where they
pan the camera around when it has a wide-angle lens on? Something strange
happens - objects come on the edge of the frame, and as they approach the
centre they simultaneoulsy shrink and slow down. As they cross the centre
and move towards the other edge, they grow and speed up. This is actually
the same problem as Brain's brick wall example, but he doesn't seem to see
it as a problem in his case.

This, of course, looks seriously wrong - we really don't see the world in
that way when we pan our own heads because, as you point out, we're looking
at an image which isn't projected on to a flat plane. You *can* fix this in
photography, by using a fisheye and projecting onto the insider surface of a
sphere - you've probably seen the amusement park panoramic cinema things
which do exactly that, but that's not a great way to view ones holiday
snaps. ;-)

>But David J. Littleboy's argument seems convincing. As he shows, the
>rectilinear lens simply must receive fewer photons from the more distant
>same-sized object, so how there could *not* be light falloff is very hard to
>understand, unless (as I suggested earlier) the aperture somehow becomes
>progressively larger for off-axis imaging.

Yes. That's one of the main objections which Brian either hasn't addressed,
or hasn't addressed very well. He's looking at this from the point of view
of the lens, without thinking too much about the nature of the world that
we're trying to take a photograph of, or about the way we see the world
directly (on that 3D surface of a sphere), rather than second hand via a
photograph.
.

Relevant Pages

  • Re: Poll on *Really* Wide Angle Lenses
    ... >>>photographic lenses are concerned. ... >> rectilinear photographic lens. ... >fisheye lens than it is to a rectilinear lens, ... Now we make a rectilinear photograph of the wall, ...
    (rec.photo.digital)
  • Re: Poll on *Really* Wide Angle Lenses
    ... >>No, BC is correct here, at least as far as flat-field rectilinear ... > rectilinear photographic lens. ... fisheye lens than it is to a rectilinear lens, ... > projection*, but that's a bug, not a feature. ...
    (rec.photo.digital)
  • Re: P&S versus DSLR -- actual photos for comparison
    ... background is blurred to impart a feeling of motion in your photos? ... The images I posted were taken at f/4, which means that the lens ... The Canon lens is not quite capable of giving me 3072 usable pixels ... Evidently you don't like to photograph moving subjects, ...
    (rec.photo.digital)
  • Re: Connecting EOS 10D to Zeiss ICM 405
    ... the distance between eyepiece/relay lens and camera ccd, ... projection distance is different from the distance it is designed for. ... looking for the most suitable camera from the outset. ...
    (sci.techniques.microscopy)
  • Re: True life size photo
    ... > how to take a photo so the resulting print image is exactly true life size. ... The easiest thing to do is to take a picture of a ruler with the lens ... enlarge/scale the photograph to match. ...
    (rec.photo.digital)