RE: Re: Sigma 10-20 any experiance?
- On Mon, 04 Dec 2006 09:32:58 -0500, Sacha Griffin wrote:
> These are all good arguments.Yes, I can see that being an issue. However, with a screen width of
> Things I've noticed that have influenced my decisions...
> 1. When shooting 10mm rectanlinear sphericals in areas of high detail (ie
> grass), its REALLY difficult to compress down under 7MB without downsampling
> and downsampling. The amount of detail even at the same resolution you would
> take with a fisheye like the sigma f4 is astounding creating a larger file
> at the same resolution.
> 2. People don't zoom. It's hard enough to get them to read any instructions
> on the screen for anything. It's quite amazing and disheartening.
1600 pixels, all you must do is to set the initial zoom field of view
to 60 degrees (hardly "zoomed in"), and you will properly sample *all*
the detail in a 10000x5000 pano without any more zooming required. In
the fullscreen era, some of the old maxims aren't as useful.
> It also depending on the subject. I like the clarity in your shot, am notI agree that for a full sphere this would probably have been about
> ignorant about zooming, and enjoy the scene. If you did this as a complete
> sphere you'd either need to downsample or living with slow delivery issues.
> For areas, where you are delivering photography for the clients of your
> clients and know you are dealing with people that most likely won't zoom or
> even click and drag despite every blinking instruction... creating sharp 360
> photography downsampled and compressed to 2-3 MB seems perfect to do the
twice as large, but a full sphere shot at 10mm (instead of 18mm) would
be about the same or even a bit smaller (5MB, say).
When delivering hundreds of sphericals is your business, every byte
counts, and it may indeed be overkill to go beyond 5000x2500, or for
that matter to use full screen display. But that is based more on
business decisions than technical decisions. It's perfectly possible
with today's technology to target ~2 arcmin/pixel (twice the "ideal" 1
arcmin/pixel I mentioned) without wasting pixels, and without
requiring the user to zoom in. In fact, I would hazard a guess that
in the pano I posted, you were already interpolating pixels on your
screen when you loaded it (primarily because I start reasonably zoomed
in). Going forward, this will only become easier (as
bandwidth/processor/graphics cards improve), and more pressing (as
monitors grow in size and pixel density).
An interesting side-question is whether and when it will be "easy" to
acquire panos at the mythical 1 arcmin/pixel. The answer comes in
considering the pixel pitch of digital camera sensors going forward.
The maximum pixel pitch of a DSLR today is roughly 180 pixels/mm
(5.5um pixels). Smaller pixels suffer greater noise (limited by
photon noise, so no way around it), and out-resolve the image circle
delivered by even very good lenses, especially at small apertures,
where diffraction dominates. Here's an interesting take:
So, at 180 pixels/mm, to obtain the "ideal" target resolution of 1
arcmin/pixel, you must shoot at:
f = 1 arcmin/pixel / 180 pixels/mm * (1 radian/(60*57.3 arcmin)) = 19mm
On a full-frame sensor, shooting at 19mm offers about 65 degrees of
view in portrait orientation. Shooting a full sphere will thus
require two or three rows, and at least 6-8 images around (and many
more on a 1.5x cropped sensor camera).
So it seems even with advances in cameras and detectors, the only way
to resolve 1 arcmin/pixel in the future in a system with single row
full 360s is to use a camera with a much larger sensors (physically),
say 70mm x 50mm. This type of sensor is pretty much guaranteed never
to show up in a consumer or mainstream pro system, simply because it
would make the entire camera much too large. Here's a 39MPix digital
back with 6.8um pixels at 50mm x 36mm. Note how large it is: