This dialog is perennial as the grass. The gamut of a scanner comes into
existence when it outputs a tagged image file.
If you want to understand why the gamut has the shape it does, you have to
look behind the scenes of the conversion of radiant flux or whatever you
want to call light, pared down into a narrow visible spectrum, emitted from
a lamp with certain properties, reflected off an object with certain
properties, sensed with an electro-optical device with filters of various
other properties, and processed by SW filters with yet further properties
to fit into the tristimulus model. There''s a lot of room for
interpretation of the meaning that data! And you can't go backwards from
the data to the scene.
But why does this situation perplex us? There's nothing more surprising
about it than how you can't work back from a photograph to all the scene
contents that were outside of the frame — if you try you will just generate
organized nonsense, without any truth value. This not only doesn't trouble
most of us, we actually rely on News from such media as expressing reality!
And the whole history of the world of entertainment could not function
without such editing.
Nonetheless, we can make some workable conjectures. By my reckoning, you
can clearly argue that the output of a scanner has a gamut. For the casual
operator, he will be satisfied when the capture is rendered on his display
in a pleasing fashion, even though how his image file relates to the
physics of the media will always be open-ended. In ordinary sense of the
tools we use, there is precious little colorimetric connection of the input
domain of the illuminated object and the image file we get. Yet somehow —
and it's not a total mystery — we can end up with results that look pretty
damn good. More surprisingly, we don't even need color in the sense of
"chroma" for it to look good. This makes a weird sort of sense from the
history of imaging. It may be a mystery that works in the same way that
when a great artist scrapes down a piece of marble into the likeness of a
god, we can feel touched by that sculpture as if it the god were alive.
Personally, I feel a giddiness at the apparent paradox that we so carefully
control the output side of the colorimetric process, while the input side
remains open. But just because it's open-ended doesn't make it arbitrary.
If you rigorously characterize and align a display, then adjust the scanner
controls to produce a life-like rendering of the object on the platen, this
works well for reproduction! OTOH, even including forum members, many
people don't want dissect and explain every feature of the capture
scenario. We can discuss the features and traits of the capture process.
The firms that design the sensors for scanners and cameras must have
detailed descriptions of the filters used in the gear. I never see these
discussed in any imaging forum, so they are either so proprietary they
can't be discussed, or so standardized that they're not worth discussing.
I feel we just have to face if we don't go into the technical details of
the sensors, we are left with semantics and occasional bickering.
To the original poster of this topic, who wanted to compare his scanner
response to a commonly used RGB space, there are some experiments you can
perform, with standard charts, to explore what you scanner does with
certain media that allow you to describe some better-than-totally-vague
characteristics of the system response and make some comparisons across
media. And the good news is you get to say what the scene to device mapping
should look like! Welcome to scanner profiling. If you were capturing an
archive of objects, you might establish a baseline characterization to
strive for consistency of measure that you could say to the viewer of the
archive, these things looked pretty much like this. Items can be compared
in those terms, even if you can never work back to the deeper nature of the
object itself. But what is this "deep nature" and where does it reside? Is
all that needs to be satisfied are our perceptions?
There's a recent trend of progress in illumination where lighting systems
are applying the same sort of lamp technologies as used in tristimulus
displays. Today, in order to use a profiling app to characterize a monitor,
I need to know the lamp technology for the backlight so that the device
primaries can be compensated for the standard-observer functions to hone
the accuracy of the alignment. If common lighting technologies also apply
tristimulus approaches, do the input and output systems play well together?
. . . If we build a world illuminated by TV, what is the fundamental nature
of that world?
Let's face that in a deep sense, the capture of image data can be nothing
more than a simulation of a field presented to the camera. You can't
rebuild the object from the scan. But you can present a very workable
visual simulation of it! It workd well because we traffick in images as
much or more than imaged objects. Maybe we ought to be amazed what people
will accept as real? Things may get super weird as 3D printing evolves.
Poynton himself regards a similar observation of TV monitor alignment in
his proposal for what became the BT.1886 specification.
Picture rendering, image state, and BT.709, C. Poyton, 2010
http://poynton.ca/notes/PU-PR-IS/Poynton-PU-PR-IS.pdf
To paraphrase him: The purpose of a standard alignment is to honor the
artistic intent of the content director; to allow the future to recover
what he saw when he signed-off on delivery of the program. This was needed
because a vast shift in display technology was causing the characteristics
of the HD CRT studio reference monitor to be forgotten, and no one had ever
bothered to specify its response in an capture-independent way!
Artistic judgement is the watchword of the process.
Per the original post, colorimetric gamut is imposed during the capture
process, and you can't work backwards from the colorimetric structure to
the field beheld by the scanner. But this shouldn't trouble you. It's how
the whole colorimetric system works!
And as more of what we see is generated automatically, there will be no
deeper world than the generator. I for one welcome our computer overlords!
I once stood upon the pyramids of Giza rendered by Ubisoft's Assassin's
Creed Odyssey.
I hear there are great advances in natural language processing, something
called GPT-3, where a computer can barf up a report than looks to most of
the world it was written by experts but is in essence just a statistically
distributed set of previously ingested lore from existing reports. I think
the CIE may have been on to something.
Don't get trapped in the equals sign. There is no equality except in the
idea of equality. The usefulness of the tristimulus theory is what it
leaves out. Although Allah guide us if in long run all we end up with is
images on TV.
On Thu, Feb 25, 2021 at 1:24 PM Gerhard Fuernkranz <nospam456@xxxxxx> wrote:
Am 25.02.21 um 16:19 schrieb Hermann-Josef Röser:
emulsionThe volume in Lab space is only relative to the material being scannedyes, that is correct. The scanner's ICC-profile only refers to the
with which the target was produced. It also only refers to theillumination
specified in the reference table, the LEDs in use, exposure time and
CCD-detectors in use during profile generation. To my understanding, all
these together define the input device.
