Add more iterations or use a wider scale or a combination of both.
1 Like
Also I think the method is intended for use in combination with filmic highlight reconstruction. I’ve managed to get quite good results when using both.
1 Like
12 seems to be the max in the module even when I try to type a higher value.
What is supposed to happen? Will the guiding laplacians method sample colors from bordering regions to use? Perhaps this question isn’t directed to you @g-man. 
I think using an image from PlayRaw might be the best option. I picked this one that has clipping at the sensor. Panasonic Reds: How to??
Here is my quick xmp edits using 3.9.0+1050~gcabcf8f2a
P1039009.RW2.xmp (10.5 KB)
The filmic6v with HR at -1.42 and wide transition to makes it decent from a distance. Zoom in is different:
Clip highlight method
GL method mid range settings
Gl max settings (12 iterations/2048pix)
I am assuming the filmic -1.42EV HR threshold is after exposure module settings (based on the location in the pipeline). This is with filmic HR threshold set to 0.
Again, I’m just trying to learn how to use this correctly.
2 Likes
I have the same difficulties especially with larger areas where all three channels are clipped.
Here is the demonstration:
This is how the scene looks with a correctly exposed photo:
Here is an overexposed one with corresponding intication:
This is how it looks without highlight reconstruction:
Highlight reconstruction with LCh.
The overexposed area of the wall under the windows is too bluish but overall relatively good reconstruction:
Reconstruction with guided laplatians all values at maximum:
Note that we still have magenta on the wall.
Guided laplatians on maximum and reconstruction in Filmic. This is the best I could do with it:
Reconstruction is a little better if internal iterations in Filmic are set to 0:
At higher values the color of the frame bleeds into the wall I guess. The color of the framework confuses the results, since it is very close to magenta:
I get the best results in this case as a combination of reconstruction in LCh and Filmic:
Here is the file if you want to try it yourself:
DSC_2319.NEF (28,6 MB)
Edit:
I just did another test, and the best result with Guided laplatians in combination with Filmic is now this:
For Guided laplatians I chose default settings again.
5 Likes
The more i think about this issue, the more i think it would be nice to make the highlight reconstruction mask in filmic (there is a button to show it) available so it can be used to tweak the colors (away from magenta).
For years I’ve wondered why highlight reconstruction doesn’t look for magenta and mitigate when it’s detected. In all my years of post processing raws it’s always been the magenta pain in the butt and I’ve never needed such a colour in a real scene. Those who photograph flowers may feel differently. It just seems like such a clear sign of error as it’s visually so obvious.
1 Like
Nice idea 
Well, there’s another problem with that module: it uses the raw white point after the white balance module. And in most cases, esp. red can have a WB coëfficient > 2. So with a 14-bit raw file, you consider a red value that uses 15 bits as clipped, where it isn’t clipped in the raw data before WB. The same can happen with the blue channel.
(Have a look at some clipped areas with WB module on and off, quite revealing…)
A lot of flowers have colours in the yellow-red-violet/magenta range. I really wouldn’t like automatic “correction” of magenta hues for my wild orchids.
3 Likes
Sometimes, we deal with chromatic aberration selectively, but I would caution the generalization. As you say, there are exceptions like flowers.
Not helpful, until you substantiate the sarcastic opinion.
Now I have a great idea. On setup you get asked “do you photograph flowers” yes, no. If no magenta will never feature in your images 
Joking aside, as a user it’s so blatantly obvious magenta doesn’t belong in the frame, its fascinating that it’s hard for a program to detect this. Perhaps it’s guided by the ambition that it should never fail and mistakenly kill magenta. Thing is it fails the other way by producing magenta all the time. Presumably orders of magnitude more often.
I mean there is object recognition but I don’t think FLOSS raw processors or its devs are ready to do or accept that. Anyway, general purpose tools are much more useful than a tool that only works sometimes.
It is so obviously not a well thought out idea that nosle himself finds fault in it literally two sentences later.
Indeed but sarcastic remarks don’t contribute to the discussion either. I prefer @rvietor’s reply.
1 Like
There are a growing number of us who are tired of the diatribe. How would you like us to express this fatigue and annoyance?
3 Likes
I think there has been a request for true sensor clipping.data …armed with that users should be able to make informed evaluation and correction as necessary ???
1 Like
You mean a hint to the degree of clipping? Please elaborate.
This just 10 hours ago is likely the followup spoken about…
How can this help? Seeing something doesn’t allow you to edit it. It seems like we diverted from the subject.
Maybe this should be a new post because it really isn’t about the guided laplacian. The real subject to me (and I think s7habo too) is how to treat the magenta highlights in darktable. A short answer could be, don’t overexpose. But some of my shots are of my kids and they move thus changing the lightning.