Hi All, There's a really good darktable blog entry 'Color Reconstruction' on the dt website from Ulrich. He goes into great detail as well as of course knowing rather a lot more about the internals than most of us.
In addition to that I'd say that a quick instant fix in many cases is to activate the 'shadows and highlights' module. If that doesn't work a simple common sense approach is to start by reducing the overall exposure, if the pixels aren't actually blown that will bring the h/l detail back and you can then play with curves/masks/whatever to get the rest of the image presented to taste. If reducing the exposure (even to extreme) doesn't bring the h/l pixels back - then I'd read the blog page and follow that approach. Good luck, Rob -----Original Message----- From: Remco Viëtor [mailto:[email protected]] Sent: 04 February 2017 13:28 To: [email protected] Subject: Re: [darktable-user] How to fix overblown white? On samedi 4 février 2017 09:03:18 CET Michael Below wrote: > Hey, > just a quick note about the basecurve: that curve is very early in the > processing stack. So if you blow your highlights in this stage, they > are gone, before you can do something about it. If there are exposure > problems it is far better to use the contrast curve in the colour tab, > because that is applied later. Cheers Michael Yes and no. Note that I said it was not very useful to disable the basecurve *if you are going to enable it again later*. In DT, modules are applied in a fixed order, independant of when you activate them. In addition, DT works internally in 32- bit floating point, and does not clamp the colour values within the pipeline (that's what the developers tell us, anyway: see section 3.2 in the manual). Of course, colour values *are* clamped to a fixed range at the end of the treatment (if only because you don't want values to wrap around when going to 8 bit/channel), but that's not relevant in selecting which modules to use. That is not to say that picking another basecurve to spare your highlights can't be useful, but this *will* influence the whole image, just like the tone curve. That was something the OP wanted to avoid. And that requirement means you will have to use some form of masking. I find that parametric masks can work rather well in such circumstances. And don't forget that, in general, you will want to apply a rather aggressive S-curve to transform your sensor data (intensity linear with photon count) to something pleasing for you (intensity ~logarithmic with photon count). You have a choice of using one of the provided basecurves (optimised to reflect the basic style of different camera brands) or build your own in the tonecurve module. Remco ____________________________________________________________________________ darktable user mailing list to unsubscribe send a mail to [email protected] ____________________________________________________________________________ darktable user mailing list to unsubscribe send a mail to [email protected]
