Or maybe the color shift is in the OOC JPEG and you got used to it ?
Be careful with shifts, they are only defined by comparison with a reference. Mixing in metamerism errors of sensors loosely respecting Luther-Ives conditions + sub-optimal 3×3 input color profiles (assuming sensor are linear when they are not really), it’s impossible to characterize a shift without a spectral ground-truth.
All you can do is enabling/disabling filmic and see if the colour looks the same at a different exposure. But comparing OOC rendering to software output makes no sense.
Also, remember input colour profiles are approximations optimized to give better colour rendition for (caucasian…) skin tones. Blue is at the other end of the chroma wheel.
So to get “good” sky colour… One would have to do different correction or what? And TBH in your (@anon41087856) example according to my wife (waaay better photographer) OOC JPEG is “worse” than edit, mainly because colour is actually better and more “real” + there’s no halo effect near mountains + the sky and sea in righthand side do not fade to gray.
Sure. Just saw the same difference on images you provided.
Okay. Will do this and let you know. But the results look really unnatural without correction. I may post it here for evaluation. Also I may provide a RAW to play with. Altitudes were up to 5000m, so the sky in reality is very dark and has violet/magenta shift. And I expect to see it after development.
…is always a challenge because the color of sky is always different. I saw sky to be yellow, orange, green… I’d agree with @anon41087856 that it’s a matter of taste and natural look instead of “right” color representation. However, there are always cases that we may say for sure that color is wrong. I’m talking about such a case.
Left is the spectral sensitivity of the CIE 2° observer, right is the spectral sensitivity of the typical camera. That huge overlapping of sensitivity means your “green” camera channel is pretty much a full-spectrum/panchromatic channel instead. Hoping you can match this RGB to XYZ by only a 3×3 matrice is delusional, you clearly see that it’s not a linearly independent (orthogonal) vector base, so your dot product means bogus. Ain’t no linear algebra in non-orthogonal vector bases, unless you extract eigenvector and make them orthogonal.
Where people can help others is good. I don’t have a Facebook account either and I generally try to inform people why it’s a bad place, but people still have their own free choice to use it or not. Let us not infringe upon the rights of others to do as they wish.
Result after having applied filmic as follows:
• Turning on filmic
• Setting the middle grey luminance to 18.5%
• Clicking the white relative eyedropper
• Clicking the black relative eyedropper
• Adjusting the middle grey luminance to 12,4%
Here is my development with the .xmp. The WB is set by the snow and corrected one click toward margenta. I did not try to fix the tint on purpose – to see how the color is rendered. As you can see, there is green tint on the sky as well as on the far moutnains. It should have proved my point, but frankly, I’m quite sure it’s a camera input profile to blame. The coefs were probably set in a wrong way for DNG… 20190228_114736.dng.xmp (8.7 KB)
@anon41087856
I have played a bit with highland shots. My current impression is that
Probably something has changed in the latest GIT version since I cannot see anymore the greenish shift I was mentioning
I’m still not happy with how the sky is rendered colorwise, but starting to think that most probably the input color profile or the camera itself may be involved, because it was Nikon J1 with quite small sensor and therefore the color rendering is far from perfect. And I had to push the Color Balance 110% in 150% out to have brown tones pop, so the sky becomes oversaturated. With more regular 100% in 125% out the sky are more natural but other colors are washed out. So it’s probably some inconsistency in input color profile coefs, not an erroneous color transformations.
And I think that without extra correction the sky is rendered lighter than it’s supposed to be on such altitudes.
On some such shots RGB (max) works better for the sky, on others Luminance Y. The latter makes sky a little bit oversaturated, though.
Here are couple of shots. No additional color correction was made beyond standard filmicRGB + Color Balance → output saturation. I think that color is more or less fine on both of them but I feel the tone should be darker.
The result is incredible. It would have been better if you edit the same picture on a different module of Dark table and also upload it here so that we can see the difference. Great effort and the picture came out nice.
Thanks a lot. The achievements in 3.0 are indeed impressive and very helpful to see it all in one well referenced place; my apologies for the oversight.
In terms of this specific simple guide I don’t think there’s really much different in filmic v4. The middle-grey slider in filmic is now disabled by default but the guide already references the new recommended way to do it (use the exposure module).
The saturation can now be tweaked in the filmic module itself but the color balance method is still a reasonable way to do it (and I’ve still found it better to use color balance on some images).
