Hi Andrew,
As I understand it, non-linear curves (S-curves) can compress light and dark tones, which can lead to a loss of detail, especially if they’re too aggressive. that’s why I’m trying to restore them after this one.
Here’s a short video demonstrating this phenomenon
Greetings,
Christian
What I find challenged about these images is, we’re not looking for objective colors. The sensor is not clipped, so what the camera recorded is likely accurate.
However, we perceive a color shift for very bright areas. This color shift is not real. It’s just something our eyes and minds make up when the cone cells start to saturate.
So what we’re trying to recreate here is a psychophysical response, not an objective reality. Therein lies the challenge of this render. It’s not about reality, but about perception.
hi Christian,
I totally agree about compressing light and dark tones, this is their job, to make a big range fit into a smaller space. But compression doesn’t have to be across the whole image, you can (should?!) identify what the photo is about, what is important, and you can increase the contrast for the subject a lot if you wish (at the expense of other parts of the image). Your example is interesting. If the intention is for the photo to be about the vehicle, then you could bring the bodywork more to the centre of the histogram and “hyundai” and other details would probably be quite visible. Or if it was about the driver responding to the other person, then you’d need to brighten up a lot and then maybe you don’t care what the manufacturer is? Either way a think filmic (or sigmoid) plus exposure can give good contrast for the subject. But I generally use some local contrast too, usually Contrast Eq.
If the subject is the whole image, perhaps a landscape, then it’s trickier.
@cedric yes all “local” areas are affected. But a given tone value is affected the same amount regardless of where it is in the image, whereas with local contrast the amount depends on what that particular bit of image is close to.
I couldn’t agree more. For me, the strength of software like Darktable is that it gives me all the tools I need to reproduce what I saw and perceived when I took the shot.
That’s precisely my point.
In the case of this photo, if I increase the exposure to emphasize the vehicle coming out of a very narrow market, I lose too much detail in the body of the truck.
It goes without saying that this is very subjective.
You say “can”, Christian but, to my simple mind, contrast is directly proportional to the slope of the transfer characteristic and, with an ‘S’ curve, highlight and shadow contrast is always reduced.
P.S. … by “contrast”, I mean the Michelson Formula (Imax-Imin)/(Imax+Imin). Pardon the pedantry.
that’s why I’m trying to restore them after this one.
Here’s a short video demonstrating this phenomenon
Sorry, couldn’t follow the video … too quick for my old brain and no narrative on my machine …
I used color equalizer to keep the render pleasing. I think in this case, the goal is not to avoid gamut clipping at any cost, but rather to obtain a good looking picture. I tried to keep the orange/red hues and to preserve minimal detail in the clipped areas.
I haven’t touched any colour related setting, besides white balance. I think the colours look similar to the ooc jpg:
I used sigmoid as I always do, and I used it as I always do with the setting rgb ratio if I change this to per channel the whole appearance changes drastically. And not for good. There is a shift towards red and the contrasts and fine nuances are lost so that the flower looks now more opaque:
Bumping this as it occurs to me it might be worth seeing how AgX handles it. Just a very quick edit.
I think clearly is more proof that AgX will solve all of our problems 
World peace, free beer, eternal life: the PBL tone mapper. Next on my list.
Can I add a feature request to fix my stomach issues?
No need, it’ll be built in.
Could you clarify if these 4 items are being completed in sequential order or whether PBL will realize the first 3 projects?
PBL will solve the first 3, plus the stomach issues, and others, too. I’m just doing the port, though; the algo comes from Chuck Norris.
The best feature is its simple and effective gamut management approach: it roundhouse-kicks out-of-gamut colours into oblivion, eliminating the issue.
