This video is a must watch

You just blew my mind! It’s a good point though, and something that I had never considered before.

Yes, it seems to be harmless most of the time. Here’s some feedback from the author ot the video on that topic:

Right, so I have the “Highlight Reconstruction” module disabled, as I use highlight reconstruction in the “filmic rgb” module, when necessary.
That “Highlight Reconstruction” module is meant for when “filmic rgb” module is not enabled.
It’s better to use highlight reconstruction tab in “filmic rgb” module, when “filmic rgb” module is enabled.
It sounds convoluted, because it is…
To be honest, I was puzzled to why it’s ON by default in the latest Darktable, it shouldn’t be, IMO.

I didn’t watch the video. What is this cyan business about?

Fun fact: my brain did not compute when I saw the word norm. Been using norms in my own processing and reading about norms in dt discussion. It is also past midnight. :full_moon_with_face:

Yeah, sorry, it was without the tone curve.

Interesting. I have not found the reconstruct tab in filmic rgb to be of any use. I have a preset that sets the Highlight reconstruction module to “reconstruct in LCh” mode… automatically and for all images. Otherwise I sometimes have pink skies.

I never do anything else with the HR module, but I make use of the scene tab in filmic. I tried turning off HR, and using only filmic and don’t always get the same results. I can’t get the reconstruct tab in filmic to do anything useful.

I understand you because I had a hard time trying to figure out the best way to recover highlights with filmic rgc.

After watching some videos of Aurélien and reading/learning about the development of the filmic rgb module in this forum I’m now have a way to recover highlights and I’m very satisfied with it.

Basically I deactivate highlight reconstruction altogether, and set preserve chrominance in filmic to ‘no’. After this I go to the reconstruct tab of filmic and adjust the threshold and maybe some of the other sliders. Part of those steps were an advice of @kofa in this thread: Lost highlights in darktable - #2 by kofa

1 Like

I don’t think I ever recommended that.
I find that with bright highlights, luminance Y works well. For my taste, ‘no’ results in too much desaturation.
After I made that comment, Aurélien developed his new guided laplacians method for highlight reconstruction. And I still use LCh from time to time.

1 Like

I edited my comment, I hope I expressed it better now :wink:

1 Like

It’s probably best to watch but he talks about how in typical RAW processing software the highlights in a blue sky shift hue into cyan before heading towards white.

Being a spry, young 30-something hue shifts in digital imagery is what I’ve seem my whole life so I expected and when filmic didn’t do it looked wrong. So much of our expectations and what looks like reality to us is baked in by our culture and the technology of the time. To this day I swear the real sky shifts to cyan near the horizon to my eye except on exceptionally clear days.

As for “the norm,” I just meant it here as “usual” or “expected.” From about my teenage years on digital filming and photography have been everywhere. Even my poor rural high school had one of those floppy disk using Sony Mavicas back then. I didn’t get a personal digital camera until the early to mid 2000s but I was still relatively young and in college at that point.

5 Likes

My point was the normalization of norm on this forum led my brain to compute it as the mathematical term, just as you tend to expect cyan.

What colour is shifting to cyan? An oddity about me is that I don’t think in colours, so I have difficulty talking about them, if that makes any sense. I need someone to describe it to me.

It is visually described in the video, quite well.

Okay. I have time right now. I will check it out!

Edit: I watched the video. What I don’t understand is the off then on filmic part. The hue shift of blue is an addition of green to make cyan before all channels are blown or too bright for our eyes to see colour. Why does the image have more green when filmic is off? Does that mean the data has higher green values relative to blues prior to the application of filmic? My thought would be that blue would remain blue unless something shifted the colour at some point. I don’t always think clearly, so someone please elaborate. Thanks.

Can this explain some of what you might see flipping filmic off and on…

2 Likes

Are you masking it…you need to drop the threshold the right amount. Its not perfect but when things are completely blown it can remove the gray or magenta issues…

@priort So, it isn’t so much a cyan issue as a clipping one, and I suppose the cyan (and the yellow in the sunset PlayRaw thread) are examples where the non-clipping green contribute too much in bright(ened) areas. If that is the case, then I understand. My approach to processing doesn’t encounter these problems normally. Or at least, I don’t notice. Ha ha.

I really don’t know if it fits this specific case but I was just throwing it out as a possible contributor. for sure though also the norm will shift the individual color channels differently and so when you have clipped and near clipped channels you can see so things happen in an exaggerated way in those areas depending on which norm is in play

Yeah, colour as we see it is more like a sponge than strict simplistic ratios. :balloon:

Ah this is the key for me.
For the set of images I am working on right now, none of the reconstruction sliders in filmic do much of anything useful, but the preserve chrominance modes filmic makes the big difference. Then just sliding the white relative exposure sliders if there is still any magenta.

Now I like the appearance with reconstruction in filmic (left snapshot) vs the highlight reconstruction mod.

1 Like

Feel free to post problematic images as PlayRaw (add a licence, e.g. Creative Commons — Attribution-ShareAlike 4.0 International — CC BY-SA 4.0, that allows others to develop your image and post the result).

My eyes open NOW!
I thought that cyan shift was normal! Holy smoke it’s not!

7 Likes