Tom Scales wrote:
> I'm guessing, but I think the theory is simple.
>
> Pixels.
>
> Since you're going to be overlapping anyway, you might as well make the
> 'non-overlap' dimension be as large as possible. Makes for more overlaps,
> but an awful lot of pixels.
Yup, that's it -- imagine the sensor was 3:1, for the sake of
argument, and covered a view 90 degrees wide by 30 degrees high.
Ignoring overlap for simplicity's sake, you have two options:
1: four photos, landscape mode
2: twelve photos, portrait mode.
Version 1 uses fewer images, but the final image is only 30 degrees high.
Version 2 uses more images, but the final image is 90 degrees high.
In general, if you've gone to the effort to set up a tripod and
special rotating plate, the extra time to take more shots is trivial on
top of that. Even hand-held, it's not a big deal taking more shots, I find.
Now, if you want to get really fancy, you can tilt the camera so that
the _long diagonal_ is vertical -- that gets you the most vertical
coverage with a single strip, though stitching can be a bit fiddlier to
work out what's going on. This is what I've been doing for fisheye
panoramas with an E-body, because that way, with 8 photos (6 if I'm
extra careful), I can pretty much cover the entire sphere; there's a
hole at top and bottom, but as long as I'm careful about what I'm
standing on top of/underneath, it's usually possible to patch that up
afterwards, and it saves the fiddling around of N/S/E/W/U/D shots where
it's a huge pain keeping my feet out of the way for the verticals.
(it's _much_ easier to do the patching if you map from cylindrical to
6-faces-of-a-cube view; pano2qtvr free edition does that, and it's very
very useful).
-- dan
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|