i wrote some code in a branch once that would move our denoising before black point subtraction (or at least the clamping to 0). the differences to the old code path in terms of result images were disappointing at best. first, the black point thing would only result in very academic differences, only noticeable in cases where the image contrast would look very flat and boring (base or tone curves adding a little toe slope to the blacks immediately make the difference disappear). also, unfortunately doing denoising in bayer instead of interpolated colours degraded the denoiser's quality a fair bit. so i deleted all the code.
that said, today's pipeline does not clamp at 0 when subtracting the black point, meaning profiled denoising will get all the input colours without clamping. the anscombe transform there does interesting things especially with tones close to the black point, so i don't think we currently have a denoising gain from this change. i have a few local commits using a fisz transform instead, which may potentially work better near the blacks. if you're really interested i might publish the code somewhere. so far i didn't see a real-life example that would justify spending a lot of time on this though. may be worth noting that we went through most of this before the opinionated discussion started on the internet. i would assume the same holds for most other software having to deal with this. cheers, jo On Mon, Oct 24, 2016 at 6:59 AM, Zbigniew <zbigniew2...@gmail.com> wrote: > On Sun, Oct 23, 2016 at 07:47:58PM +0200, Jean-Luc Coulon (f5ibh) wrote: > >> Do you mean that you HAVE HAD the problem you report or is it simply "to >> know if" ? > > I'd like to learn it, since I'm considering a switch from LR. But I don't > know Darktable yet, therefore got no idea whether does it suffer the same > problem. The simplest way is to ask them who know this program much better, > isn't it? > > So you don't know the answer, it seems. Anyone else? > -- > regards, > Zbigniew > ___________________________________________________________________________ > darktable developer mailing list > to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org > ___________________________________________________________________________ darktable developer mailing list to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org
