Re: [Gimp-developer] GIMP useability - choosing linear vs perceptually uniform RGB
- From: Elle Stone <ellestone ninedegreesbelow com>
- To: Gimp-developer <gimp-developer-list gnome org>
- Subject: Re: [Gimp-developer] GIMP useability - choosing linear vs perceptually uniform RGB
- Date: Sat, 02 May 2015 12:40:16 -0400
On 05/02/2015 10:21 AM, Gez wrote:
El sáb, 02-05-2015 a las 06:09 -0400, Elle Stone escribió:
But you're not proposing to add a toggle to gradients alone, you're
proposing to put them*everywhere*.
Yes.
And your reason is that users have to decide how operations are
performed, no matter the result, no matter if it makes any sense or it
doesn't.
In the specific case of perceptual vs linear RGB, the user is the only
person qualified to say whether it makes sense or not.
The toggle or other mechanism for allowing the user choice between
perceptual and linear RGB should plainly indicate the default choice
made by the devs. But the user needs the option to override the default
choice made by the devs.
A tooltip that educates the user about why the default makes sense would
be a nice addition.
So what about other aspects like alpha association then? Should toggles
be added everywhere for that too so the users can decide if alpha will
be associated or unassociated?
Well, you might be able to answer that question. I'm not qualified.
Personally I don't use alpha channels except in the extremely rare
instance when I'm exporting a png with a transparent background for use
on a website.
I'd like to see this discussion heading towards a real world list of
examples of real needs for such options that can't be satisfied with
anything else than these toggles.
You are presupposing that the devs can foresee every possible use to
which a user might put a given editing operation.
No, I'm saying that users like us should describe real world situations
where certain options are needed in order to convince developers of the
necessity of such options.
"Let me do whatever I want" is not a good argument.
Gez, you want explicit, known-in-advance justifications for every
exception to your proposed rule that all RGB editing operations should
operate on linear RGB, with no user choise to override the default.
Let's say that as developers we take the time to examine every single
RGB editing operation supplied by GIMP. Let's say we can find good
reasons for allowing 50% of all RGB operations to have a user choice
between linear and perceptually uniform RGB:
First, how much is the UI less complicated if only 50% (or 30%, or
whatever percentage you'd like to pull out of a hat) of the RGB editing
operations have a means to allow the user to override the developer-set
default?
Second, what if we make a mistake with one of the "no user option"
operations? Right now drawing a gradient can only be done using
perceptual RGB, and on radiometric grounds that is a mistake. But on
artistic grounds, it depends on the artist's intention. Giving the user
choice with respect to linear vs perceptual RGB provides a necessary
safety net for developer mistakes.
Unlike Robert Krawitz's wonderful examples from Gutenprint, GIMP users
aren't running any risk of physically harming their digital darkroom
when they go against developer defaults. The worse thing that might
happen is they might learn something about why certain defaults were
chosen in the first place.
The need of linear and perceptually uniform gradients is a real need.
You can easily document when you need one or the other and create simple
examples.
Now, give me a good example why scaling should be better done in
perceptual gamma (other than preserving legacy appearance, which is the
ugly situation that took us here in the first place).
Personally I can only think of two specific use cases:
Pedagogic: teaching a class on why scaling should be done using linear RGB.
Artistic: I was quite amused to see that the lights in this image
twinkle as the image is scaled larger and smaller by the Firefox
browser: http://ninedegreesbelow.com/tcb/img/pine-branch-over-lake.jpg
I can imagine an artist deliberately making a background image where
lights twinkle as the image scales up and down. You will rightly say
that there are other ways to acheive the same effect. So what?
You'll find soon that aside from keeping legacy appearance,
> the situations where you need operations to actually work in
> perceptual gamma are rare.
I don't have a personal stake in keeping legacy appearance, not having
any legacy files. Nonetheless, removing user choice regarding perceptual
vs linear will invalidate an awful lot of user workflows right off the
bat before the user has a chance to figure out why linear gamma RGB is
often a better choice to get them to where they want to go.
I've been working with linear gamma image editing in the specific
context of editing high bit depth photographic images since around 2005,
which doesn't make me an expert, just saying I do have some experience.
I prefer to edit using linear gamma RGB as much as possible. However,
the instances where I choose to use perceptual RGB aren't rare but
rather fairly common depending on the particular editing task and
artistic goal.
There are specific editing operations for which I always use perceptual
RGB. And there are other editing operations for which I usually use
linear RGB, but sometimes use perceptual for particular artistic reasons.
So in practice, combining linear and perceptual back and forth during
your work is not something you need all the time.
Do you perhaps mean that combining linear and perceptual is something
that *you* don't find a use for very often? Because I do precisely
switch back and forth quite a lot.
Tell me for instance why in your UI proposal you merged a layer using
the screen blending mode in perceptual gamma.
What's the need there, what's the effect achieved?
In this particular case screening perceptually uniform instead of linear
RGB produces a brighter image for use as the base layer when blending
the blurred image.
I could have acheived approximately the same effect using linear RGB for
Screen, followed by Curves. But Curves isn't so easy to "program" for
shadows when operating on linear RGB, because the points on the curve
end up being very close together.
Why should the user be forced to resort to Curves to modify the results
of a blend mode performed on linear RGB, if what she wants is what she
would get automatically by using perceptual RGB?
I don't like it and I'd prefer that a true linear workflow is
implemented where nothing has to be flipped to perceptual unless there's
a good reason. And I bet that those good reasons would be rare, real
exceptions that could be treated as such.
When working on any given photograph, I do find good reasons to switch
from linear to perceptual and back fairly often, even though it means
doing an ICC profile conversion because I disabled the babl flips before
compiling babl/GEGL/GIMP, because the current UI is confusing to the
point of being unuseable. In my opinion. How do you get linear RGB for
Curves? How do you get Perceptual RGB for Filter/Noise/Add RGB noise?
The babl flips can be used either to *enable* or to severely *restrict*
the user in one very specific area, which is using linear vs
perceptually uniform RGB. I hope the devs choose to enable the user.
Best,
Elle
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]