PanoTools:
Re: dynamic range and noise (yet again)
pedro_silva58 2005-Oct-28 13:26:33
greetings!
the direct inspiration for my post came from the idea recently
floating here to read out succesive frames from the sensor during a
single exposure. but instead of reading them in a ramp up as usual
(eg 1/500s, 1/250s, 1/125s), i thought that taking more frames at say
1/500s might have some advantages. i wasn't sure it made any sense,
so decided to run it by you. i thank all who wrote in reply to my post.
i had written:
> here's an idea, related to dynamic range and noise. like bracketing,
> it would only work for static subjects. ...
> we take, say 8 shots with that exposure value (with tripod, remote,
> etc).
"Erik Krause" wrote:
> However, the technique has it's limits: Camera shake, moving objects
> and so on.
i thought i had said that much...
"Erik Krause" wrote:
> You just re-invented a commonly used technology to reduce noise
> by avaraging several shots. Helmut Dersch wrote PTAverage for that
> purpose: http://www.path.unimelb.edu.au/~dersch/average/average.html
JD Smith wrote:
> 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.
different people chose to reply to different aspects of my idea (which
is normal). i knew i was not inventing anything new, but had not seen
the combination of adding and averaging before -- adding the shadows
(to build up exposure), averaging the highlights (to lower noise), and
using both for the midtones (to get both effects in this important
area). i thought this combination might have some merit.
"Erik Krause" wrote:
> Hint.: Adding up the images would of course increase the exposure but
> since you easily can cross the 8 or 16 bit limit if you don't use a
> floating point format ...
JD Smith wrote:
> Note that you would need to perform this averaging
> in a higher bit depth space, preferrably floating point, to achieve
> the best results.
so, it may be a good idea to do both adding and averaging in a higher
bit space than the original and final images. good point i hadn't
thought about.
"Paul D. DeRocco" wrote:
> Mathematically, the sum of a bunch of short exposures has the same
amount of
> noise as one longer exposure. ...
> I think that if that provided a measurable benefit, then cameras would
> already be designed to do that automatically for long exposures. All
these
> tricks are certainly well understood by the designers of sensors and
digital
> cameras.
are you sure the d2x doesn't do it? no, not automatically, for (like
noise reduction for long exposures) this sort of thing is not always
appropriate, but...
try multiple exposure: with gain=on it averages up to 10 new
exposures, with gain=off it adds them. overlay does the same thing,
to files already on the card. the good designers at nikon must know
what they are doing...
hugh crawford wrote:
> As far as I know the random noise is where there are pixels that are
> brighter than they should be ( thermal noise AKA Johnson-Nyquist
noise) and
> thermal noise never causes pixels to be darker. What I would do is
take a
> group of exposures and discard all pixel values that are greater
than the
> median, and then average the values that are less than the median.
...
here's an interesting idea by hugh, probably worthy of further
consideration by the local experts. anyone?
cheers,
pedro
------------------------ Yahoo! Groups Sponsor --------------------~-->
Fair play? Video games influencing politics. Click and talk back!
http://us.click.yahoo.com/T8sf5C/tzNLAA/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/