Topics

[cml-post-vfx-aces] Color gamuts

deanan@gmail.com
 

There's often a bump in red sensitivity around 400nm (from 350ish-425ish) on many sensors. 
Where the IR(and UV) cut filter turns on affects how much of the red crosstalk you get in blue, which is more apparent in underwater especially after you try to white balance by increasing the red channel. 

Deanan DaSilva
Consultant
Playa del Rey,CA


On Wed, Jan 31, 2018 at 12:17 AM, <daveblackham@...> wrote:
A side issue but may be a pointer to the cyan question. This is more a comment from a practical perspective than a scientific test. We get issues replicating cyan a lot the time when filming underwater, its an unforgiving environment for cyan and blue so we see the problem a lot, usually grading them we end up with Magenta blooming errors being produced, it can look like magenta cast to the entire image. Using format -20 UVLAR filter help which is a UV cut filter to remove any unwanted UV. 

Gavin Greenwalt
 

> I do find it interesting how such mathematical gamuts are used for marketing purposes. For example, RED Wide Gamut and Canon's color gamut show a green that's fantastically far out.

They aren't for marketing purposes, they're to avoid hue shifts etc when people apply contrast curves and other color corrections.  The dirty secret of RGB color science is that, while RGB triangles are useful and they're easy to transform on a computer using basic math, they don't actually describe color accurately and all of our color correction operations are hacks.   Color gamuts of course aren't a looping circle with some physical light spectrum of 'magenta' between blue and red for instance, it's just a convenient way to represent a mix of red and blue light as our eyes perceive it.   

So why move the primaries off to the side? If you have a color like
ACES RGB(0.5,0.5,0.5) [Hue: 0]   
...and you apply a 2.2 gamma to it you end up with...
ACES RGB(0.72,0.72,0.72) [Hue: 0]
That's a 'safe' transformation, since Red Green and Blue came in as the same value, they came out the same hue. 

If though you had a color like:
ACES RGB (0.7, 0.3, 0.2) [Hue: 12]
...and did the same simple mathematical operation to decrease contrast you end up with...
ACES RGB (0.85, 0.57, 0.48)  [Hue: 16] 
... a slight hue shift.  For comparison If I do the same operation in RedWideGammut though the hue shifts to 14 (+2 shift vs +4 in ACES). 

Neither ACES AP0 nor Red nor Canon are "right" they all have unintentional hue shifts from basic color operations but they all balance different tradeoffs. 

Gavin Greenwalt
VFX Supervisor
Straightface Studios

On Wed, Jan 31, 2018 at 3:37 PM, deanan@... <deanan@...> wrote:
There's often a bump in red sensitivity around 400nm (from 350ish-425ish) on many sensors. 
Where the IR(and UV) cut filter turns on affects how much of the red crosstalk you get in blue, which is more apparent in underwater especially after you try to white balance by increasing the red channel. 

Deanan DaSilva
Consultant
Playa del Rey,CA


On Wed, Jan 31, 2018 at 12:17 AM, <daveblackham@...> wrote:
A side issue but may be a pointer to the cyan question. This is more a comment from a practical perspective than a scientific test. We get issues replicating cyan a lot the time when filming underwater, its an unforgiving environment for cyan and blue so we see the problem a lot, usually grading them we end up with Magenta blooming errors being produced, it can look like magenta cast to the entire image. Using format -20 UVLAR filter help which is a UV cut filter to remove any unwanted UV. 

JD Houston
 


On Feb 2, 2018, at 11:37 AM, Gavin Greenwalt <im.thatoneguy@...> wrote:

they don't actually describe color accurately and all of our color correction operations are hacks.   

+1


Neither ACES AP0 nor Red nor Canon are "right" they all have unintentional hue shifts from basic color operations but they all balance different tradeoffs. 

A linear RGB space is NEVER going to match what the eye sees - eyes are non-linear both in a small area and across a field of view. Hue angles also shift just because of brightness changes
(which is not modeled in L*a*b*)

So the convenient linear color spaces all are going for different things based on where they place primaries. The design process is complicated been more
when you consider HDR/WCG performance.

The cameras don’t really have detected colors in those triangles that go all the way out.  But pixel values can end up there as it is the encoding space.

Jim Houston
Starwatcher Digital
Pasadena, CA

Art Adams
 


They aren't for marketing purposes

Well... when I look at representations of RED Wide Gamut and see how exaggerated the green primary is, they don't explain that this is a mathematical construction designed around reducing hue shifts. No one does, but that one example seems a bit extreme and it certainly had me wondering. It's definitely being used in several places as a marketing tool, but I'm sure its primary function is not that.

Thanks, Gavin and Jim, I'm learning a ton. I know this because I'm now convinced that everything I know is wrong. :)

--
Art Adams
Director of Photography
San Francisco Bay Area

JD Houston
 


On Feb 2, 2018, at 6:36 PM, Art Adams <art.cml.only@...> wrote:

I'm now convinced that everything I know is wrong

Always  a sign of progress…

The more we know, the less we understand. ;-)

Jim Houston
Pasadena, CA


Argyris_Theos_cml
 

Come on now!
That's my heritage!

"I know one thing, that I know nothing"
Socrates
5th century B.C. 

Argyris Theos, gsc 
DoP, Athens Greece,
+306944725315
Skype Argyris.Theos
via iPhone

3 Φεβ 2018, 8:00 π.μ., ο/η "JD Houston" <jdhouston@...> έγραψε:

I'm now convinced that everything I know is wrong

Always  a sign of progress…

The more we know, the less we understand. ;-)