On Sat, Dec 22, 2012 at 7:12 AM, Matt Feifarek <[email protected]> wrote:
> On Thu, Dec 20, 2012 at 3:00 PM, johannes hanika <[email protected]> wrote:
>>
>> cool, pushed that to master. let me know if your presets disappear (in
>> which case i did it wrong).
>
>
> They work; it's auto-detected. Now that I can run the Dt branch with this
> feature, should I do the profile step, or do partial ISO settings, or
> anything else?
>
> It works astonishingly well! I noticed that it doesn't do well with high-iso
> chroma noise... leaves behind some colored patchy blotches (like noticed on
> the original blog post). Is there anything I can do to make that work
> better? Perhaps the original photos were too monochromatic or something?

well, there's the style refered to in the blog post, to remove them
after the fact with the equalizer. i'll experiment a little with a
different distance metric (ignoring color) and maybe one more paper
for adaptive search radii (increasing that globally will alleviate
that matter, but become ridiculously slow. maybe we can also just
expose it so the choice is yours).


> I'd also be interested in advice for how to manipulate the settings to
> retrieve some of the lost small-grain details (where possible). I realize
> that at some point, small details will look like noise to the algorithm, and
> the "signal" and the "noise" are very close, and therefore cannot be saved.
> But any tips for trying to retrieve/rebuild thin lines for example ... that
> would be super helpful.

make the patch radius bigger (all the way to 4 if you want), it will
match fewer patches and thus become sharper. you may have to increase
strength to counterbalance it, or else the denoising effect might be
weakened too much. i might also revive an old branch that was using
multiple high-iso shots together to denoise one shot, and see if that
helps in this framework. but that'll work with the same profiling
data, so that part of the work should be done.

> This is great stuff guys, I am super impressed and really grateful.

:)
-jo

> -- Matt

------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
darktable-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/darktable-devel

Reply via email to