|
|
Weeeeeell... Now that I've stuck my head back in here, you all know
there is NO WAY that I can NOT chime in on this thread, even if I'm a
little late to the party... so, where to start?
Am 07.04.2021 um 13:26 schrieb Kenneth:
> Well! Now we're getting somewhere. There *is* a major misunderstanding, although
> I'm not yet sure where. Here's my own breakdown and logic:
I guess I see where you're coming from. Here's the kicker: When we're
talking about "linear color", what does "linear" even refer to?
"Linear" is always in reference to _something_, and when talking about
brightness (of colour components or otherwise), that something could be
either:
(1) physiological stimulus of the human visual system; as in, "this grey
looks half as bright to me as that white piece of paper over there";
(2) raw color values; as in, "those are the numbers that Photoshop gives
me";
(3) input signal for a classic CRT display; as in, "this is what I get
if I set the input signal to 50mV";
-OR-
(4) physical light intensity, as in "this LED emits 1 mW of red light".
Most people never come into close contact with (3) or (4), and it so
happens that (1) and (2) [in conventional computer graphics] are roughly
proportional to each other. So for people dabbling with computer
graphics, it is easy to conceptualize (1) and (2) as being exactly the
same, and "the" reference system for brightness.
Even once you realize that (3) and (4) are also a thing, and that (1)
and (2) may be quite dissimilar in modern computer graphics, it is easy
to remain stuck in the concept of either (1) or (2) as being "the real
deal".
In my experience, this is something you just have to snap out of, to get
a true grasp of all the gamma shenanigans.
When people like Ive or me talk about "linear color values", we ALWAYS
mean "linear with respect to PHYSICAL LIGHT INTENSITY".
There's one important reason why this is the "frame of reference" we're
using: It is the ONLY one that is TRULY UNIVERSALLY FIXED.
(3) is an unstable reference, because no CRT (or any other analog)
display responds to the same input signals in exactly the same manner.
This is even true for CRTs of the same brand and model.
(2) is an unstable reference as well, because there are different
standards as to what a value of e.g. 128 out of 255 really is supposed
to indicate; that's what you see when
Even (1) is an unstable reference - possibly even the worst of all, as
the human response to visual stimulus varies not only between people
(think color blindness as an extreme case), but also depending on
ambient viewing conditions, and whether you happen to have stared at a
red or blue piece of cardboard just 20 seconds earlier.
Trying to get a grip on gamma while holding on to the color values in
Photoshop as your frame of reference is like trying to build a mental
picture of the solar system while holding on to a geocentric world
model: It might seem intuitive, but you'll have to wreck your brain with
complex and weird epicycle stuff, only to leave you with a bizarre
picture that, although possibly of some practical use, remains
incomprehensible at its core.
You have to smash that idea out of your head, and start fresh, putting
the physical light intensity smack bang at the center of your world
model, as the figurative sun.
Once you've done that, you can look back at good old Earth (read:
Photoshop color values), and see how - and why - that is itself a moving
target, just like all other pieces in the puzzle that is gamma handling.
(And once you've gotten there, you'll realize that there's a whole
galaxy out there, and that even "linear RGB values" in the physical
light intensity sense aren't really a static thing either. Welcome to
the world of metamerism, color models, gamuts, dynamic ranges, and
what-have-yous.)
Well, I'm starting(?) to ramble. The takeaway message is this:
Photoshop color values are NOT THE LINEAR THING. They're fluid. They're
subject to interpretation. They're just binary codes - bit patterns that
have no meaning until some meaning is assigned to them. Even the numeric
representation presented by Photoshop should be considered little more
than a courtesy for humans, so that they don't have to memorize long
sequences of ones and zeros.
"50% Grey" is NOT THE LINEAR THING either (unless in a program
implemented by Ive, of couse ;) ). It is fluid. It is subject to
individual perception, surrounding colors, and what you've stared at for
breakfast. And don't you dare call it "middle grey", or you'll start a
flame war of epic proportions.
THE TRUE LINEAR THING is the stream of photons that your screen emits
when it displays an image. Or the stream of photons that enters a camera
when it takes a picture. Physical light intensity.
Unfortunately, physical light intensity isn't trivial to produce or
measure reliably. However, there's one potentially eye-opening
experiment that should work reasonably well with most displays:
- Start Photoshop or the like, and create a new image with exactly the
same number of pixels as your display.
- Painstakingly paint the entire image with alternating rows of black
and white pixels. Make sure the white rows are truly white, and the
black rows are truly black.
- Display the image in fullscreen mode.
There. You've turned half of your display's pixels to 100% brightness,
and the other half to 0% brightness. The average physical light
intensity of your display is now exactly(*) the arithmetic mean of the
physical light intensities of your display's white and your display's
"black".
(*Well, roughly, to be fair. Depending on display type, there might be
some slight "bleeding" effects between pixels that can skew the result a
bit.)
Now you can draw rectangles of uniform grey color onto that image, and
see which one - when squinting your eyes - best blends with the striped
background. The color of THAT rectangle you're seeing is true 50%
physical light intensity grey. (And no, those values Photoshop is
showing you, they ARE NOT that color. It is additional context that
under certain circumstances makes them CORRESPOND TO that color.)
Once you've accepted Physical Light Intensity as your One And Only
God... oh wait, no, that's the wrong text...
... anyway, once you've gotten to THAT point, you can actually start
unraveling mysteries like "then what do the Photoshop values even mean?"
In a nutshell, those "Photoshop values" boil down to, "well, we want our
images to be not just black and white, so we'll add 254 distinct steps
of brightness in between. Somewhere. Wherever. Doesn't matter much, as
long as the hardware is cheap to build. It's not like anyone would ever
want to do _computations_ on images, right? That would be ridiculous.
We'll worry about precise specifications later, when everyone has gotten
accustomed to it and thinks there's any system to this madness."
It was only when that madness had blown up in people's faces that they
started thinking about standardizing those brightness steps, and some
de-facto standards emerged from the ashes. Multiple different ones, of
course, because a single one would have made life too easy.
You can choose which of these standards Photoshop should follow, by
changing the gamma settings. By tweaking this setting, you change which
set of Photoshop values CORRESPONDS TO a certain color.
In a similar vein, you can look at all the other color-related
components in the whole system: The color value system used by POV-Ray's
parser, the color value system used in the render engine, the color
value system used in file formats, and actually a host of other little
puzzle pieces - all of which can only be brought together properly if
you use physical light intensity as the one immovable fulcrum of the
whole thing.
There. I guess I've rambled enough for a single post. More thoughts on
this thread later.
Post a reply to this message
|
|