Re: [Gimp-developer] Attempt to summarize the discussion of my examples of what doesn't work in unbounded sRGB

[Elle Stone <ellestone ninedegreesbelow com>]

On 04/21/2014 04:45 PM, Øyvind Kolås wrote:
> On Mon, Apr 21, 2014 at 1:16 PM, Elle Stone
> <ellestone ninedegreesbelow com> wrote:
> > On 04/21/2014 07:07 AM, Teo Mazars wrote:
> > > Hmm, I understand, then the default black-to-white gradient would be non
> > > perceptually linear, which is more surprising than the color-to-color
> > > gradient. I think I am now convinced this is correct, but it will probably
> > > be puzzling to use.
> > >
> > > BTW, "gradient" is not such a good example because it's not related to
> > > chromaticities, "Invert" would be a better one.
> >
> > I don't understand what you are trying to say. How is drawing a gradient
> > from most saturated red to most saturated green in any given color space not
> > related to chromaticities?
>
> Picking the gradient color stops (including start and end) colors can
> be considered as separate from the color space the rendering process
> and interpolation is done in.
>
> > The color space chromaticities are what determine the most saturated
> > possible colors in any given color space.
> >
> > Aren't the negative RGB values required to draw a gradient in unbounded mode
> > sRGB from Rec. 2020 reddest red to Rec. 2020 greenest green exactly what
> > makes the gradient so horribly wrong when drawn in perceptual rather than
> > linear? It's the effect of the perceptual TRC that sends the "out of bounded
> > sRGB gamut" colors to strange places.
>
> This might be an argument that do the _interpolation_ by default in
> for instance CIE Lab instead of sRGB.

No. It's an argument to make gradients using linear light, not perceptual.

Drawing a gradient works just fine in unbounded mode sRGB as long as you 
convert the image to a true linear gamma sRGB, and most likely also just 
fine in default GIMP if gradients were able to be drawn using linear light.

I can't test that because for some reason someone decided that the right 
way to draw a gradient was to use perceptual instead of linear, for both 
linear and gamma precision. But in my modified babl/gegl/GIMP that has 
the babl conversions back and forth between linear and regular TRC 
disabled, in a true linear gamma color space drawing a gradient is one 
of the things that does actually work in unbounded mode sRGB.

On another note, if the answer to "it doesn't work in unbounded mode 
sRGB" is "do it in CIELAB instead", then you are turning GIMP into a LAB 
editor. LAB is a wonderful adjunct to RGB editing. But operations in LAB 
should never be a wholesale replacement for any RGB operation.

If drawing a gradient didn't work in unbounded mode sRGB (but it does if 
linear light is used) the answer wouldn't be "use CIELAB". The answer 
would be "use the source color space primaries, not the sRGB primaries".

GIMP is the best by far RGB image editor available to open source. 
Depriving GIMP of *any* RGB operation by replacing it with a LAB 
operation is a move in the wrong direction. *Adding* LAB functionality 
is a very good thing. *Replacing* RGB functionality with a LAB operation 
weakens the usefulness of GIMP as an RGB editor. Things mix differently 
in LAB. Color is XYZ/RGB first. LAB is derivative.

Elle