PanoTools:
Re: dynamic range and noise (yet again)
JD Smith 2005-Oct-27 18:33:57
On Thu, 27 Oct 2005 16:04:49 +0000, pedro_silva58 wrote:
> greetings, all!
>
> here's an idea, related to dynamic range and noise. like bracketing,
> it would only work for static subjects.
>
> if this makes any sense, it shouldn't be too hard to implement in,
> say, photoshop, or a small dedicated program.
>
> imagine we expose for the highlights, ie, measure exposure so as to
> take the highest important highlight as close to clipping as possible
> without actually clipping. we take, say 8 shots with that exposure
> value (with tripod, remote, etc). if the scene dynamic range is high,
> shadows will be severely underexposed, and mids somewhat underexposed,
> too.
>
> now, we open all 8 images, and add (photoshop-screen?) them together
> -- call it image S. this will increase the exposure the equivalent of
> 3 stops, lightening the dark parts (opening up the shadows). next, we
> do the same with 4 images only -- in fact, do it twice, once with the
> first 4 images, then with the next 4, then average them together, and
> call the average M. then, we also average all 8 images, and call that
> average H.
>
> image S should have an okay exposure in the shadows, possibly with
> less noise than a single long exposure. image M should be okay for
> the midtones, with less noise than a single longish exposure. and
> image H will have the same exposure as the original shots, but much
> lower noise.
>
> finally we composite the three images S, M and H, pretty much as if
> they had been bracketed. S will contribute the shadows, M the
> midtones, H the highlights.
>
> this is a lot more work than a simple 2 or 3 bracket sequence. what
> do we gain?
> - long exposures are prone to noise. adding several shorter exposures
> helps (this sort of thing is often used in astrophoto for the same
> purpose).
> - with bracketing, we have a single (somewhat noisy) highlight
> exposure. this way, we can considerably reduce noise in highlights too.
> - the less noise we have, the more we can sharpen, etc.
> - with the extra work, comes extra flexibility.
> - all cameras are limited in bracketing range. this way, the sky
> (your card memory?) is the limit.
> - it is possible to combine the ideas of range expansion and noise
> reduction (eg, take 8, average in pairs, then add the 4 averages: this
> would increase exposure 2 stops, and still reduce noise more than a
> single exposure for double the time).
>
> so... does any of this make any sense?
Coadding exposures to beat down noise is a tried and true method used
in astronomical imaging. Your suggested method would amount to adding
all 8 images together with the same fixed scaling. I think you'd
accomplish the same thing by averaging 8 (ideally raw) exposures, and
then using standard contrast techniques to bring out shadow detail,
which will suffer from much less noise than a single image would have.
That said, digital photos of dark scenes are typically not
"background-limited", i.e. the noise is dominated by dark current and
read noise in the detector and read-out electronics, rather than
photon noise in the scene itself. In this regime, taking two
exposures of time T/2 and adding them together will not lead to the
same noise characteristics of taking one exposure of time T. Of
course, if you are limited by highlight clipping, T/2 may be
preferable to T. Note that you would need to perform this averaging
in a higher bit depth space, preferrably floating point, to achieve
the best results.
JD
------------------------ Yahoo! Groups Sponsor --------------------~-->
Most low income households are not online. Help bridge the digital divide today!
http://us.click.yahoo.com/cd_AJB/QnQLAA/TtwFAA/.Cr1lB/TM
--------------------------------------------------------------------~->
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/PanoTools/
<*> To unsubscribe from this group, send an email to:
#removed#
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/