Do you prefer something like that ? Patches are aligned on their values.
4 Likes
What if you were to add a similar scale along the color chart showing the middle zero point for saturation along the abscissa and for brilliance along the ordinate?
Then in the narrative you could explain the axis rotation for chroma, lightness and luminance?
v2
Notice: luminance is a physical metric, not a perceptual one. Thatās the change you get by varying exposure. All the other metrics are perceptual and derivated from JzAzBz space.
14 Likes
Bingo!
While Iām on a pretty chart spree, these are sRGB gamut volume hue slices on lightness/chroma planes (with uniform steps), for the R, G, B primaries of the sRGB space. You see the gamut lower envelope follows lines of equal saturation. That also explains why increasing chroma (along the horizontal direction) sucks : some hue/lightness coordinates will be pushed out of gamut much faster than others.
Pushing brilliance drives color toward the primaries of the space.
4 Likes
What helped me most was one of the charts youāve posted, with a flower-like bundle of coloured rays; how changing saturation maintains the distance from the black point, but varies the angle enclosed with the neutral (vertical black ā white) line; how changing chroma moves away or closer to that axis, parallel with the x plane; and how changing brilliance means moving along those constant saturation lines, closer to or away from black.
4 Likes
@kofa Like this ?
9 Likes
Yes, something like that. I think itād make sense the make the āsaturationā arrow curved (it is my understanding that distance from the black point is constant). I also liked the wording āThe lines of equal saturation are the oblique ones, so the saturation setting opens or closes their angle like a flower, while purity [brilliance] moves along those oblique lines.ā I just like the mental image it conjures up. 
The specific post was this: Editing moments with darktable - #207 by aurelienpierre
1 Like
Well, the saturation goes straight, itās not bended. We donāt really rotate around black. The original angle defines the initial orientation for both vectors, but they donāt rotate across the transform.
These are excellent!
Sentences like these help my understanding a lot.
(I did read that the second sentence is not correct. What I mean is that this style of description is helpful to me: āX changes Y but keeps a constant distance for Zā āX keeps Y constant but changes Z towards Aā)
Except that AurĆ©lien says itās not the case:
Which I have to admit I do not understand. It I take this image:
Then the āsaturation lineā (brilliance axis?) at the ādottedā end of the arrow encloses a different (smaller) angle with the vertical lightness axis than the other two āsaturation linesā it crosses. And the arrow is perpendicular to the original āsaturation lineā; if it is always (āstep-wiseā) perpendicular, then, by definition, we get a circle, do we not? If you mean that increasing saturation by, say, 10%, moves along the perpendicular vector (denoted by the arrow) a given number of units, then correct, itās not a pure rotation and the length of the vector from the black point to the new value does change, but should that really be the case, instead of something like this (sorry, the drawing is not very accurate):
Interesting aspect. I think in @anon41087856ās chart he starts at a given colour (zero point of brilliance / saturation arrows). If, starting from this coulour you increase saturation you donāt follow a circle, but the saturation arrow. This would mean that you could not reach the resulting colour in a circle but in some sort of trajectory from the starting point. Nevertheless saturation increases whith the angle of the āsaturation lineā, which fits to the āflowerā-modell.
These diagrams/discussions have now been incorporated into the development version of the darktable user manual. Any feedback is much appreciated.
One change in the GUI of color balance RGB I would like to see is to bold or in some other manner make the names global offset, etc., stand out from luminance, hue, etc. Part of my inability to find them is unfamiliarity with the module, of course, but I have a hard time finding them each time I use the module.
Yes I agree. You can add the following to the CSS tweaks section in preferences>general to make the text white:
#section_label { color: white; }
1 Like
Iāve raised a Pull Request for section labels to have the same colour as the module control labels. Theyāre already a slightly bolder font so hopefully that should be enough.
3 Likes
These dotted lines in this plot style imply convergence at 0,0,0 [0-255 range] so you assume that leads to a rotation around 0,0,0. Ask yourself what chromaticity 1,0,0 should have: Some value close to achromatic? Or some value close to the red-primary?
The dotted lines are actually parallel and thus do not converge at āblackā (remember: black doesnāt exist) so there is no rotation. What looks like a convergence at zero is down toā¦good questionā¦? Iāll go out on a limb and say ālimited precision to represent saturation changes at very low intensitiesā.
Iād love to hear other wordings and explanations on that.
(Just to be clear: there is absolutely nothing wrong with the plot and with Aureliens explanations, itās just down to interpreting the graphs right, which I admit is intricate.)
and
ā¦where the lines are not parallel needs a little more explanation. Could you explain a bit more what you mean ? At the moment Iām āstanding on the hoseā and my laborious sorted marbles are falling apart. 