You have been wisely advised by Erik and Willy. Making three or four
shots gives a better quality... but the 2-shots solution has some
Take four images at 90 degrees one from each other (but with the NPP
set as it would be set for 180 degrees spacing i.e. beta= 90 deg. See
and optimize by using
control points only for subjects that are sufficiently far (>2 m)
from the camera. Make sure that CP are also requested for at least
one of the opposite pairs of hemispherical images. After this
optimization, remove the other two opposite images from the PTGui
project and keep only the remaining pair. Save the script as a
template for further use.
Use the most precise rotator for the two-shots and a mounting
interface for the camera that is accurately repeatable.
You may know that this 2-shots technique with a fisheye is under
these conditions, the only way to guarantee absolute absence of
parallax errors along the (sole) seam whatever is the distance of the
subject! And this is why you may use a template with confidence :
Trying to make a template with a single pair of image only would be
vain: you need to have control points all over the images to get
valid correction correction coefficients (a, b, c) and offset
parameters (d and e).
Note: I have not yet found why PTGui sometimes randomly finds
successfully some control points on the overlapping ring on a single
pair of images. But this is very rare indeed;-)
My recent experiment have shown that the real useful angle of view of
the FC-E8 fisheye converter is about 190 degrees and not 183 degrees
(as it was initially specified for the old CoolPix series) when
mounted on the recent CP P5100 that has a larger sensor and most
probably a revised lens design.
The adapter from Agnos that fits the FC-E8 is too long to yield the
largest angle of view: the rear of the fisheye converter should
ideally be much closer to the front of the zoom lens when the latter
is in "Wide" and "Converter" mode settings.
I have therefore devised a way to benefit from this untold feature
and wrote an article where the mechanical data are given for the
shaving operation of the adapter: http://michel.thoby.free.fr/DuoPix/
The image is largest in diameter and is also then sharper near the
bluish blurry ring and the latter is thinner as the vignetting is
reduced. The TCA is very slightly increased but it still can be well
and simply corrected.
I presume that the Nikon UR-E20 adapter is also not adapted (!)
because being also too long but I am still waiting for delivery from
Nikon France to make sure;-(
Le 2 mai 08 à 15:12, Erik Krause a écrit :
> > I dont think that Ptgui can stitch 2 FE.
> It can stitch, but only with manual alignment. It most probably won't
> find control points automatically. But there is a chance that one can
> find some manually. There might be a problem blending with such small
> overlap with enblend, smartblend or PTGui, but there still is the
> feathering algorithm of PTStitcher and the possibility to output
> layered PSD and blend manually in photoshop.
> > Make 3 or 4 shots. You will get a better quality and no seams.
> Doubtlessly true...
[Non-text portions of this message have been removed]