Maker Pro
Maker Pro

Realistic colors for monitors ?

S

Skybuck Flying

Jan 1, 1970
0
Hello,

The RGB color space does not contain/display all colors we humans see in
reality...

In reality we humans see more colors in real-life than our monitors can
display.

I was wondering if nvidia is researching graphics cards and/or monitor which
try to display "realistic colors" ?!?

If not it would seem to me that "they" have the most to gain from such
technology and should therefore research/develop it ?!?! ;) :)

Might be the next big thing or maybe not...

I do wonder what "changes" would be needed to the current system...

How many bits would be necessary to display all colors ? Would it still be
an RGB system or would it need something else ? ;)

I can vagely remember one company investigating such technology and working
together with some animation studio... like pixar ? or industrial light and
magic ?

Bye,
Skybuck.
 
S

Skybuck Flying

Jan 1, 1970
0
Norman Peelman said:
In reality our monitors can output more colors than our eyes can
process.

Maybe it's not about "number of colors" but the "color range" itself.

Like deep black, and very white.

And very red and very blue, and very pink, very orange and so forth.

Monitors seem to be limited to a certain color range.

Bye,
Skybuck.
 
T

Thomas Richter

Jan 1, 1970
0
Skybuck said:
I was wondering if nvidia is researching graphics cards and/or monitor which
try to display "realistic colors" ?!?

For that you would need monitors with an infinite number of primaries.
As (most) monitors and the color system of the computer-monitor
connection (say VGA and DVI) are based on three primaries, all colors
that can be reproduced by a monitor are contained in a triangle in the
xy color space whose edges are given by the (phyiscal) colors of the
monitor (inside the full gammut). However, the gammut of visible colors
in this space is certainly *not* triangular, thus necessarily colors are
missing.

It is not a matter of the monitor or the graphics card, but the whole
system to signal colors.
How many bits would be necessary to display all colors ? Would it still be
an RGB system or would it need something else ? ;)

Not a matter of bit-count (alone). The bit-count only defines the
precision by which colors can be represented, not the size of the gammut.
I can vagely remember one company investigating such technology and working
together with some animation studio... like pixar ? or industrial light and
magic ?

OpenEXR, by ILM. But this is on high-dynamic range images, i.e.
representing several magnitudes of luminance. But it is still based on
three primaries, and so are (most) capture devices and (most) display
devices. IIRC, the primaries can be specified, thus it is possible to
describe "virtual" colors outside of the visible gammut and thus
describe all visible colors. However, since the monitor and the
monitor-computer link is constrained to "physical" colors, that itself
doesn't buy you much; it is a win for processing images - i.e. in the
image or movie processing toolchain, because coding loss can be avoided.
This is what it has been designed for.

Greetings,
Thomas
 
S

Skybuck Flying

Jan 1, 1970
0
Thomas Richter said:
For that you would need monitors with an infinite number of primaries. As
(most) monitors and the color system of the computer-monitor connection
(say VGA and DVI) are based on three primaries, all colors that can be
reproduced by a monitor are contained in a triangle in the xy color space
whose edges are given by the (phyiscal) colors of the monitor (inside the
full gammut). However, the gammut of visible colors in this space is
certainly *not* triangular, thus necessarily colors are missing.

I have seen drawings trying to explain it... the "true" gammut seems a bit
wobbly...

Maybe a (cubic?) spline system is needed... where the coordinates form some
kind of spline to specify the resulting color...

Would that help ? ;)

Bye,
Skybuck ;) :)
 
P

Paul Keinanen

Jan 1, 1970
0
For that you would need monitors with an infinite number of primaries.
As (most) monitors and the color system of the computer-monitor
connection (say VGA and DVI) are based on three primaries, all colors
that can be reproduced by a monitor are contained in a triangle in the
xy color space whose edges are given by the (phyiscal) colors of the
monitor (inside the full gammut).

Using readily available high efficiency CRT phosphors, the gamut is
quite limited http://en.wikipedia.org/wiki/SRGB_color_space which is
used on many computer monitors and HDTV displays. The current SDTV
primaries do not differ much from these.

The original (1953) NTSC primaries had the green point much closer to
the top of the gamut, thus capable of producing deeper greens.
Unfortunately, the available phosphors had a low efficiency, thus,
producing dim pictures.
However, the gammut of visible colors
in this space is certainly *not* triangular, thus necessarily colors are
missing.

It is not a matter of the monitor or the graphics card, but the whole
system to signal colors.


Not a matter of bit-count (alone). The bit-count only defines the
precision by which colors can be represented, not the size of the gammut.


OpenEXR, by ILM. But this is on high-dynamic range images, i.e.
representing several magnitudes of luminance. But it is still based on
three primaries, and so are (most) capture devices and (most) display
devices. IIRC, the primaries can be specified, thus it is possible to
describe "virtual" colors outside of the visible gammut and thus
describe all visible colors.

http://en.wikipedia.org/wiki/Imaginary_color

A much more radical color space is at
http://en.wikipedia.org/wiki/File:CIE1931_rgxy.png which would be OK
for a transmission and production standard, but would cause a lot of
problems displaying it.
However, since the monitor and the
monitor-computer link is constrained to "physical" colors, that itself
doesn't buy you much; it is a win for processing images - i.e. in the
image or movie processing toolchain, because coding loss can be avoided.
This is what it has been designed for.


The http://commons.wikimedia.org/wiki/File:CIExy1931_CIERGB.png could
be quite good if monochromatic emitters at 700, 545 and 380 nm would
be available. Wouldn't adding a fourth emitter at 515-520 nm cover the
whole gamut ?

While in the past it has been necessary to specify the update rate,
spatial resolution, gamma and color space according to the available
display technology (CRT), does it make any sense to design new image
processing, storage and distribution standards according to some
obsolete display standards, when most likely, the display technology
will change every few years ?
 
D

Don Klipstein

Jan 1, 1970
0
In reality our monitors can output more colors than our eyes can process.

However, usual monitors cannot output some of the colors that human
vision can process, including some that are common to see in real life,
such as:

* LEDs of a more pure red shade, such as most having nominal peak
wavelength of 660 nm. Their "dominant wavelength" (which is a color
specification largely meaning "hue") is usually around 640 or in the 640's
of nm.

* Other more-pure reds, such as common red diode lasers (usually in or
near the 645-650 nm range), He-Ne lasers (632.8 nm), and incandescent red
traffic signals (dominant wavelength is often close to 635 nm). Or any
incandescent or daylight light source (or any of most xenon light sources)
filtered by a #29 or #92 Wratten filter or Schott or similar longpass
glass filters with cutoff wavelength (50% point) 620-665 nm, or other
similar deep red longpass filters.

* Deeper non-yellowish and less-yellowish greens, and deeper bluish
greens, such as 532.8 nm green lasers, and most InGaN green LEDs including
most LED green traffic signals. And especially an InGaN green or
blue-green LED with dominant wavelength anywhere from 490 to 535 nm, after
being filtered by common deep green acrylic sheet such as green
"plexiglas".

* Some deep blue light sources, such as most common turquisish blue InGaN
LEDs with dominant wavelength 470-475 nm, 473 nm DPSS lasers, or almost
any InGaN blue LED filtered by almost any deep blue filter.

* Deep violetish-blue light sources, such as a "BLB" blacklight fluorescent
lamp, a mercury vapor lamp filtered by a deep blue filter such as Wratten
47B or a deep blue dichroic filter, many "royal blue" LEDs especially if
filtered by a deep blue filter, a violet or UV LED whose visible content
is passed through a deep blue filter, a Cree "standard blue" LED filtered
by a deep blue filter, or a He-Cd laser.

* Deep violets and purples, such as combined output of a "royal blue" LED
and output of a red one whose peak wavelength is 660 nm.

* Most CRT monitors do not show truly deep greens, yellows, oranges or
reds of any hue, since the red and green phosphors in those are only
something like around 95% saturated (much less still for green according
to the 1931 CIE chromaticity diagram, but I think largely because the
green area is "distended" in that one.)

- Don Klipstein ([email protected])
 
M

Miles Bader

Jan 1, 1970
0
Paul Keinanen said:
While in the past it has been necessary to specify the update rate,
spatial resolution, gamma and color space according to the available
display technology (CRT), does it make any sense to design new image
processing, storage and distribution standards according to some
obsolete display standards, when most likely, the display technology
will change every few years ?

OpenEXR, at least, seems more than flexible enough to handle future changes:

+ In dynamic range and precision, of course, it far exceeds available
display technology (using the default 16-bit floating-point channel
representation; for crazy applications where that's not enough, it
_also_ allows other representations like 32-bit floating-point, etc...)

+ It allows an arbitrary number of channels to be encoded

+ For the "standard" set of RGB channels, it has a way of specifying the
chromaticities of the primaries (and has explicitly defined defaults)

[I wish more software would support openexr well; if you can handle the
increased size of image files -- it's probably the best image format
around.]

-miles
 
Top