Otto PETER : 20 December 2002 11:29 wrote:
> The triangles
> you mention are similar on the object and on the image side of
the
lens,
> the image triangle is just a rotated and scaled down version of
the
> object triangle. Consider that all windows are in the same object
plane,
> and hence have the same distance from the film plane (not the
same
> distance from the camera, of course).
>
> If, however, you tilt the camera, you will get different sizes
for top
> and ground level windows (and - if memory serves, and if I did my
> calculations were correct - horizontal and vertical scale factors
depend
> on partly different parameters). You will have great difficulty
to
> correct this, since there is no simple tool to scale an image
with a
> gradient to the scale factor, at least I know of none in PS.
From: Paul Lawrence
Thanks Otto,
I should have got a ruler and graph paper out before jumping to my
own
assumptions! I must confess my comments were based in part on a
magazine
feature illustrated with two pictures of a church tower one shot on
a
view camera using a wide angle lens and one tilting a digital
camera and
correcting in Photoshop. What I remember seeing in those
illustrations
must then have been distortion caused by the wide angle lens or
perhaps
was an accurate record of the visual effect of the camera's close
viewpoint. The writer of that feature argued that digital and
Photoshop
was better than using the rising front because it didn't stretch
the top
part of the tower!
I think there should be an accurate way to achieve the correction
but I
am beginning to feel a little more comfortable with having to
stretch
the image till it looks about right in Photoshop.
Hi Paul,
A late answer - I was wrong about the various distortion tools: They do
simulate the effect of a tilted camera, and therefore are also capable
of the inverse. So far, I have used them only for minor corrections, so
I did not see that tapering along one axis also gradually modifies the
scale factor along the other axis. Thanks for the nudge to test the
stuff. I'd still love to see the ultimate un-distort-tool: Define two
lines that should be parallel (and optionally also parallel to an image
edge), two other lines that should also be parallel, but at 90 degrees
to the first pair, and a slider to dial in deliberate
under-compensation.
Otto
--
//_ Dr. Otto PETER Institute for Applied Physics
.0^0__ University of Innsbruck
/_ \ / Technikerstr. 25
\____/^ A-6020 Innsbruck, Austria
_/|_ Tel.: +43-512-507-6408, Fax: +43-512-507-2922
Email: [EMAIL PROTECTED]
===============================================================
GO TO http://www.prodig.org for ~ GUIDELINES ~ un/SUBSCRIBING ~ ITEMS for SALE
