David Rosen wrote in message <67qp26$hq6$1@Usenet.Logical.NET>...
>Alan Chung wrote:
>: Mr 645 wrote:

>: > I have a Sigma 18-35 mm zoom and it controls barrel distortion very well. I
>: > prefer to shoot it about f8 for sharpest image quality, but even for interiors,
>: > lines stay straight.

>: I also have a Sigma 18-35mm but I beg to differ from the above
>: findings. I have shot a fair number of indoor group photos with it at
>: 18mm, f8 (and also stopped down futher). A number of the photographs
>: show so much distortion at the edges that people in the group
>: photographs at the edge looks like they have gain an instant 50 pounds -
>: I mean the face are flatten, widen and stretched.

D. R. (of the Rosen variety...;-):

> Au contrair my friend..... your very accurate
> description is not of curvlinear distortion
> but a complaint of the increase in peripheral
> magnification that is the cure for curvlinear
> distortion. Mathematically, there is no way
> to represent the 3-D space seen by the lens,
> without curving some straight edges of the
> subject, unless the image magnification is
> increased continuously away from the center.
>
> Before ultra [and ultra-ultra] wide lenses
> were developed, it was felt the maintenance
> of striaght edges looked truer to the subject,
> a more accurate rendering of how we understood
> the subject. Unfortunately this is unfaithful
> to how we *see* the subject. But it became a
> standard design parameter to avoid curvlinear
> "distortion" of the *rendering-of-the-subject*
> in designing lenses. Thus "distortion" came to
> mean "curvlinear distortion in general photo
> parlance.
>
> A problem arises that defeating *curvlinear*
> distortion may be faithful to what we know
> about the physical structure of the subject
> object[s], but is quite unfaithful to true
> perspective. "Classical" perspective is quite
> false to real vision, maintaining "true" and
> parallel verticals while admitting diminishing
> image size only for horizontal displacement
> along the axis of vision. This means when
> drawing classical perspective, more distant
> objects are rendered smaller, but with a flaw.
>
> Assume a flat landscape crowded with identical
> 40 ft tall pillars. Imagine it in classical
> perspective. Of course the more distant each
> pillar is, the shorter and narrower it looks,
> scaled down but shaped the same as the nearer
> pillars. So far, not too bad. Now lets hit
> the flaw in this perspective device.
>
> We are square-on to a rank of pillars about 80
> ft from left to right across our vision. We
> stand only 40 ft from the ran. The pillars at
> the left and right of the rank are drawn as not
> being smaller than those in the center, while
> more distant ranks of pillars [seen through the
> front rank] are drawn as appearing smaller.
>
> The problem is that we are not 40 ft from the
> front rank, but 40 ft from its central pillar,
> but we are 56 ft from the left or right pillar
> pillar in the same rank ! Yet, classical style
> of perspective falsifies vision to draw pillars
> of identical apparent size at the center and at
> both ends [and all across] the front rank. If
> put another rank of pillars beyond the front
> rank, a parallel rank more ditant by 16 ft, we
> draw the pillars of the second rank as smaller
> because they are further away.
>
> The central pillar of the second rank is 56 ft
> away from us, just like the left of right end
> pillar of the front rank. Why is the central
> pillar of the second rank [like it whole rank]
> drawn to appear smaller than the end pillars of
> first rank when all three are 56 ft from the
> viewer ? It is done to match what we *know*
> about the subject, not to match what we see.
>
> Here's the coffin nail of classical perspective.
> Consider the pillars in every rank, receding to
> the distance, each rank looking smaller, the
> whole landscape looking "undistorted". Lets go
> back to the central pillar of the front rank.
> Being a tall true cylinder, it is drawn as being
> the same apparent width, top to bottom, but the
> top is 40 ft in the air and our eye is at 5 ft
> above the ground. The top of this pillar is 53
> ft away will the point directly before the eye
> of the viewer is only 40 ft away. Why is the
> cylindrical pillar drawn as having the same
> apparent width, top to bottom ? See above.]
>
> Fact is, fisheye lenses render images far less
> compromised in terms of accurate perspective in
> a visual sense. If your ultrawide lens has no
> curvlinear distortion, then the 150 lb person
> on the edge of the group is broadened to look
> like the 250 lb person in the center for the
> same reasons that eurocentric world maps show
> Greenland as being the same size as Brazil.
>
> To satisfy what we *know* about the subject we
> expect lens designers to eliminate curvlinear
> "distortion". We want photos to match classic
> perspective because we "know" that to be the
> "accurate" way to render a scene. We ask lens
> designers to falsify the reality of vision to
> match a reality that exists only in the mind.
> The odd thing is, they did it. The trick of
> increasing magnification away from the center
> [or lens axis] was their tool, the long hidden
> compromise that was not all that visible until
> we had lenses with 90 dgree of more angle of
> view, where off center [off axis] is *way* off
> off center and the trick, or compromise is now
> a form of distortion in own right, sometimes
> called "eggheaded" distortion for the effect
> it renders on heads of persons pictured near
> the edge of the frame [especially the corners].
>
> Back to the original discussion, if we use std
> photo parlance, especially concerning wide lens
> coverage, "distortion" means bowing out of the
> straight edges of subjects [barrel distortion].
> Therefore, to complain that an 18mm lens has a
> distortion problem because the edge persons in
> group shots are fattened is tabboo. The less
> [curvlinear] "distortion" this 18mm has, the
> fatter the edge folks are going to get ! Of
> course fisheye lens perspective isn't perfect
> either, but it has less distortion than the
> usual ideal of a "perfect" lens.
>
> It's all in the eternal conflict of what we
> see, vs what we know resulting in a third
> "reality" of what we *observe*.
>
> David Rosen, Ph.K.* golem@capital.net
> golem@acmenet.net
>
> * Philosopher King 3:-)


Nice explanation!!! (There is a related piece
["On Seeing and Perspective"] on my web page
under "I babble", but it doesn't attack the
issue as directly as yours does.) Nice work!
--
David Ruether
http://www.fcinet.com/ruether
ruether@fcinet.com