Introducing a new color balance mode in darktable

[anon41087856]

@Elle

My sample image was a red apple, shot raw, no other processing in darktable beyond the bare minimum to interpolate the raw file, and specifically no “make it pretty” options such as the base curve or sharpening, and using the default dcraw camera input profile.

What clipping ? Gamut or dynamic range ?

@Elle

Would it be possible to add some sort of saturation mask to the algorithm to help reign in any otherwise resulting out of gamut/clipped channel values when using the saturation slider?

darktable has that already within the parametric blending options.

@Elle

I think darktable allows adding a mask as a separate step, but I was thinking about something like a checkbox to apply or not apply an automatically generated saturation mask. There are ways to construct a pretty good sat mask using just RGB, but I don’t remember the procedure - an internet search would probably turn up the calculations.

I think what you are looking for is the vibrance module, it does exactly that.

@phweyland

If I understand properly, to see the output of unbreak I choose the ouput linear way, while to see the output of color balance (or any other classical dt module) I choose the gamma way. Is that correct ?

As a temporary hack, yes, but my fix to get the true readings has been merged in master today so you don’t need to do that anymore, as long as you don’t use LittleCMS, the readings will be right in both cases. That one with LCMS is harder to fix.

[Elle (Elle Stone)]

@gez

You can grade on Log too with benefits compared to grading display-ready transfers where the tone distribution is intended for viewing and not for processing.

What does grading involve? I think by grading you mean applying a look to the scene-referred image/footage? Perhaps using a LUT?

And I think you are saying grading is done?/is best done? on log-encoded data rather than linearized data?

What does “display-ready transfers” mean? I think you are talking about taking the already-graded image/footage and applying a LUT to map the graded output to make it suitable for display on a preselected display device?

[anon41087856]

grading is that powerful technic : http://juanmelara.com.au/blog/the-summer-blockbuster-colour-grading-tutorial

[agriggio (Alberto)]

Hi @anon41087856, thanks for the link! Very informative. I think it would be very interesting to see how this translates into a darktable workflow using the new tools that you are developing. A little demo would be a great way of highlighting the advantages over a more “traditional” (“display referred”? “wrong”?, I don’t know how to call it actually approach.

For my part, I tried to work with the tools that I know, which is essentially just RawTherapee. Here’s my way of trying to emulate what is shown in the tutorial you posted (just to be clear: without any ambition of being comparable!):

And here are the before and after pictures. The raw file is at http://rawtherapee.com/shared/test_images/woodstock_people.pef. I think it should be free to use, it’s part of the RT test images – if not, I hope @Morgan_Hardwood will clarify.

• Before:
  

  

• After:
  

  

What would be the darktable way? Thanks!

[Morgan_Hardwood]

@agriggio very good video! Demonstrates how to use the new tool in an easy way using a real-life example.

Note that the video wouldn’t play the way it is embedded in the post. I had to right-click on it, “Copy video address”, then go to that address.

I took the photo and feel free to use it for open-source testing purposes.

[agriggio (Alberto)]

Thanks for confirming!

@Morgan_Hardwood

Note that the video wouldn’t play the way it is embedded in the post

do you know the proper way of doing that? I simply added the link, pixls did the rest… but if there’s a trick to make it work, let me know and I’ll edit the post.

[Morgan_Hardwood]

@agriggio

do you know the proper way of doing that?

I think you did it properly, the forum software just failed to work properly.

[anon41087856]

@agriggio

I think it would be very interesting to see how this translates into a darktable workflow using the new tools that you are developing

It translates in darktable straightforward, the only difference is the 2.5 D color wheels of Resolve are 3 × 1 D sliders in darktable. I hade even embedded a default style to produte just that teal-orange look.

[agriggio (Alberto)]

hi,

@agriggio

A little demo would be a great way of highlighting the advantages over a more “traditional” approach.

I’d still love to see it in action… do you have examples that show to the uneducated (that would be me) the benefits? I’m trying to understand. I know you posted some examples earlier (or maybe in a previous thread) but iirc there was no raw/source pic attached to try things out. thanks in advance!

[anon41087856]

@phweyland, are you using colord or xatom to fetch the system display profile ? I think I have spotted a bug with xatom that applies the gamma correction of the display twice on the same picture. Using colord, I still get wrong color picker values (but that’s expected, and I know why - bad but under control), but the image looks as it should. Can you reproduce ?

@agriggio sorry, no time for that at this point, feature freeze is coming, and big colorimetric errors need to be fixed asap. Critical stuff (at least for me).

[agriggio (Alberto)]

no worries, I understand. fwiw, I just implemented a basic log compression in (my copy of) RT, so I can experiment while trying to understand what’s going on… I’ll report here when I have something if anyone is interested

[phweyland]

@anon41087856

colord or xatom to fetch the system display profile ?

Sorry, I don’t know and I cannot check for few days. Not sure that I can choose on Windows, but in any case I’ve kept the default value. I’ve tried to select the specific monitor profile but dt doesn’t retain it from one session to the other.

[age]

@anon41087856

No. That would break the compatibility with old edits. It has to be another module.

I’m working with Troy Sobotka on something great with log and filmic curves packed together, sit tight and give me some time

Some little info ?

[age]

@Elle

You can grade on Log too with benefits compared to grading display-ready transfers where the tone distribution is intended for viewing and not for processing.

What does grading involve? I think by grading you mean applying a look to the scene-referred image/footage? Perhaps using a LUT?

And I think you are saying grading is done?/is best done? on log-encoded data rather than linearized data?

What does “display-ready transfers” mean? I think you are talking about taking the already-graded image/footage and applying a LUT to map the graded output to make it suitable for display on a preselected display device?

Grading could be just the simply rgb curves or brightness/contrast sliders.

I interpret it in this way:
“You can grade on Log too with benefits compared to grading rec709 gamma transfers where the tone distribution is intended for viewing and not for processing( with a base curve and a lot of post processing like jpegs from camera).”

[patdavid]

@Morgan_Hardwood, @agriggio - I’ve manually just dragged and dropped the video into @agriggio’s post so the file is hosted here now and should embed fine.

I’ll investigate why it didn’t fetch it.

[anon41087856]

What about some results ?

This is with default orange-teal colour-grading I have added as a preset in the color balance, + the filmic module I have developped these last few days with the help of Troy Sobotka (who is now estranged from here and yet was one of the few helpful, accurate and relevant contributors).

[agriggio (Alberto)]

@anon41087856, really good!
looking forward to trying this out

[Carmelo_DrRaw]

@anon41087856
I have started to code a color correction module in photoflow, based on the ASC-CDL formulas, with the idea of making it compatible with the DT tool (I.e. same parameters = same output).
However, the DT code seems to deviate from m the recommended implementation in at least two places:

• if I understand correctly, the RGB parameters are always re-normalized to keep the luminance unchanged. Is that correct? Would it makes sense to make this optional?
• looking at the code it seems to me that the saturation step is applied before the SOP. However, all the documents I have found recommend the saturation adjustment to be the last one. Maybe I have mis-interpreted the code?

[hanatos (jo)]

oh wow. if that is the case i think we very much want to get things straight before a release

in either case can’t hurt to discuss this exactly. can you point us exactly to the text of the spec and the lines of code you think are not in line?

[Carmelo_DrRaw]

I am mostly referring to this post, where Troy mentions:
“Also note that blender does not provide functionality to affect saturation as part of the CDL node.To control saturation on scene referred data a custom node group needs to be created to operate after the CDL node. Maybe a topic for a new question”

Conversely, the saturation is applied at the beginning of the operation in DT: darktable/colorbalance.c at master · darktable-org/darktable · GitHub

Here are a couple of examples of the re-normalization of RGB SOP parameters:

darktable-org/darktable/blob/master/src/iop/colorbalance.c#L1498

darktable-org/darktable/blob/master/src/iop/colorbalance.c#L2047

Hope this helps…

EDIT: another document describing the reference implementation: Digital Cinematography: Fundamentals, Tools, Techniques, and Workflows - David Stump - Google Libri