haze-and-glare-mountains-2023-10-08-14-02-25_P1010723_01.RW2.xmp (24.0 KB)
dt 4.4.2
2 Likes
Thank you for the play; tricky, very tricky.
I also “cheated” - in my case using a tinted graduated density.
2023-10-08-14-02-25_P1010723.RW2.xmp (10.5 KB)
1 Like
I think this recovery of the blown part of the sky is remarkable and way better than the attempts made in Darktable. Is it possible to ‘steal’ this feature and implement it in Darktable? (Not for me, for the developing team)
Well, it’s not exactly recovery; I kind of cheated, using the spot removal tool to clone in adjacent parts of the sky. Darktable already has a spot removal tool, I think, and a very good one for that matter. Thanks, anyway 
1 Like
Yes, the spot removal tool in DT is good. But the tool in RT to fill surroundings into a blown spot is better than e.g. found in FilmicRGB - imo. That’s the reason for my wishful thinking.
There was no recovery magic involved. An area near the top-left corner was cloned into (or was used to heal) the blown area. RT did not do any “fill surroundings into a blown spot” automatically.
Spot removal on (from @ThomasM’s version):
Spot removal off:
Here is what was cloned:
1 Like
Here is a version where I used the same method as what @ThomasM did, using approximately the same area. I processed the photo with the current development preview (built from master), using tone equalizer and sigmoid (with a smooth-like preset but with sRGB as the space), with a strong highlight saturation boost with color balance rgb placed after sigmoid, as suggested in https://discuss.pixls.us/t/introducing-primaries-feature-for-sigmoid/40450:
2023-10-08-14-02-25_P1010723.RW2.xmp (23.0 KB)
5 Likes
A usable result with cloning, and a very nice result with the ‘Thomas method’.
Still the efficiency of the fill-in (from the surroundings) in Filmic RGB would hardly suffer from being as good as the one in RT (imho).
You have mentioned the filmic highlight reconstruction several times. However, it is not really reconstruction:
- it does not work with raw data
- has no idea what pixels are clipped on the sensor
- instead of hard (sensor-level) clipping, it works with the white level set in filmic
- its purpose is not really to recover detail, but to smoothen the transition to white.
Darktable has recently received a rather powerful highlight recovery method, but even that cannot deal with the situation where large areas have all channels clipped (unless there are smooth gradients it can propagate into the blown area – but there will be no pixel-level detail, only smooth propagation of lightness and a colour. Check out this thread for (early) examples from Iain Fergusson, who came up with the methods, which were then later made available in darktable as the inpaint opposed and segmentation based methods.
In this photo, it would have had to re-synthesise texture, which it cannot do.
1 Like
But RT doesn’t have an automated fill-in tool does it? I thought @ThomasM just used the clone/spot removal tool, the same as @kofa did in darktable?
Yes, because a number of sliders related to the ‘enable highlight reconstruction’ option in Filmic RGB seem to work in that direction per intention, but without really offering an obvious color propagation like the one in Rawtherapee. I prefer to work in Darktable, so I would welcome an improvement in that module or area. Further being a photographer with no understanding of what moves under the hood, I just express wishful thinking. Is Filmic RGB perhaps a AuPi-module, and nobody really wants to try to improve it?
Filmic is, in fact, a module by Aurélien. It has been updated several times since he left the project; his last contribution was on 24 April. History for src/iop/filmicrgb.c - darktable-org/darktable · GitHub
Please do not invent conspiracy theories.
Once again, Filmic’s highlight reconstruction is not about recovering areas clipped at the sensor; it’s about the transition between the user-selected white point and the surrounding areas. It simply cannot do what RawTherapee’s colour propagation does.
The colour propagation in RawTherapee is much better than what darktable had for a long time. That is no longer the case since darktable has inpaint opposed and segmentation based methods, which propagate colours as well as gradients.
As far as I know, there is no version processed in ART or RawTherapee that uses any kind of highlight recovery that was more powerful than what darktable offers. Or can you point one out?
What?
To be found in which modules?
The new default (since 4.2 iirc) is inpaint opposed, which is a vast improvement on the older options.
Segmentation further improves on it in some cases, at the slight cost of some manual input.
I’d as far as to say that it renders the filmic reconstruction unnecessary, except in very occasional and specific cases. That was always meant to be used in conjunction with the guided laplacians mode in the highlight reconstruction module (above) anyway, which was VERY slow.
Thanks, Steven.
I have fumbled a lot with the parameters in ‘highlight reconstruction’ with the inpaint opposed method without getting a satisfying result. Likewise with the sliders in filmic rgb - in both cases without experiencing the same pleasant result as from RT.
For a lot of other reasons I prefer to edit my raws in DT.
Of course with this I am hoping for some creative developer to look into what the differences behind the curtain could be between RT and DT in this isolated area - which ‘secret’ is making the difference (?)
1 Like
Can you point me at a photo that worked better in RT?
MMmmm. A while back there was some discussion around this point, as before dt got the newer options in highlight reconstruction, RT did have an edge, at least in ease of use.
The general feeling now is that dt matches or slightly exceeds RT’s capabilies in this area now - having said that, different images, different styles etc, will probably give different results.
As @kofa said above, if you could provide an example where you find RT better, along with your results in dt, it would help us come up with bright ideas 
2 Likes
