Pointer Gamut Coverage Reconciliation

[blue-j]

I've compared Colour's output for Pointer Gamut coverage using the Monte Carlo method to what looks like a vigorous study carried out by Simon Baker at TFT Central: https://tftcentral.co.uk/articles/pointers_gamut

Some of the deltas are concerning. Of course we have to check if it's user (me!) error first. Here's an example of my calculation of Adobe RGB (1998):

from colour.models import RGB_COLOURSPACE_ADOBE_RGB1998 as A98
prng = np.random.RandomState(2)
colour.RGB_colourspace_pointer_gamut_coverage_MonteCarlo(A98, 10e3, random_state=prng)

95.13590844062946

Here is the comparison between Colour and Baker:

Color Space	COLOUR Pointer Coverage	BAKER Pointer Coverage	Delta
sRGB	82.07%	69.40%	12.67%
Adobe RGB (1998)	95.14%	86.20%	8.96%
Adobe Wide Gamut RGB	95.37%	99.10%	-3.73%
DCI-P3	88.30%	86.90%	1.40%
NTSC (1953)	98.15%	89.50%	8.65%
ProPhoto RGB	96.06%	100%	-3.94%
Rec.2020	99.93%	99.90%	0.03%
ACES RGB	100%	100%	0%

I welcome anyone's feedback on this! Unfortunately, Baker doesn't articulate his method of calculating the volume. He may be doing it in 2D, since he's so focused on chromaticities.

• J

[KelSolaar]

Yes it is hard to say how his computations were done, we do them in 3D, I would recommend that you adapt the colourspaces to the Pointer's Illuminant SC.

[blue-j]

SIgh. I cannot figure out how to adapt an entire RGB colorspace to a different white point. In reviewing some of my code over here I see also I did not intelligently adapt XYZ values to a different white point either. can you point me in the right direction? i'm not lazy and value your time very much! - J

[tjdcs]

It might be helpful to read SMPTE RP177, which can basically help you compute a transformation matrix from one RGB system to another (including balancing for a different whitepoint) this is basically the math behind derived NPMs in RGBColourspace.

But strictly speaking, it is not adaptation which is a physiological "white balancing" in the retina. This is best done using a "chromatic adaptation model" of which there are several in colour. CAT16 is exceedingly popular.

[blue-j]

i can do all of that for an array of values, but for a colorspace to a different white point... no entiendo the exact syntax to be used. I cant crack it

[blue-j]

well, i think i figured out how to adapt an XYZ value to a different white point? an example going from D50 -> D65:

XYZ = np.array([0.20654008, 0.12197225, 0.05136952])
XYZ_w = colour.TVS_ILLUMINANTS['CIE 1931 2 Degree Standard Observer']['D65']
XYZ_wr = colour.TVS_ILLUMINANTS['CIE 1931 2 Degree Standard Observer']['D50']
transform = "Bradford"
XYZ_2 = colour.adaptation.chromatic_adaptation_VonKries(XYZ, XYZ_w, XYZ_wr, transform)

print(XYZ_2)

[KelSolaar]

There is a dedicated method to adapt a RGB colourspace: https://colour.readthedocs.io/en/develop/generated/colour.RGB_Colourspace.html#colour.RGB_Colourspace.chromatically_adapt

Your above code looks correct by the way!

[blue-j]

import numpy as np
import colour
import warnings

# SUPPRESS WARNINGS
warnings.filterwarnings("ignore")

prng = np.random.RandomState(2)

# chromaticity coordinates for SC light source white point
w_t = np.array([0.31005673, 0.3161457])

for colorspaceid in colour.RGB_COLOURSPACES.keys():
    rd = np.array(colour.RGB_COLOURSPACES[colorspaceid].primaries[0])
    gr = np.array(colour.RGB_COLOURSPACES[colorspaceid].primaries[1])
    bl = np.array(colour.RGB_COLOURSPACES[colorspaceid].primaries[2])
    p = np.array([rd, gr, bl])
    w_r = colour.RGB_COLOURSPACES[colorspaceid].whitepoint
    w_r_name = str(colour.RGB_COLOURSPACES[colorspaceid].whitepoint_name)

    colorspace = colour.RGB_Colourspace(colorspaceid, p, w_t, w_r_name)

    rgb_adapted = colorspace.chromatically_adapt(w_r, "SC", "CAT16")

    pointer_percentage = colour.RGB_colourspace_pointer_gamut_coverage_MonteCarlo(
        rgb_adapted, 10e3, random_state=prng
    )
    print(colorspaceid, ": ", pointer_percentage)

Seems to work fine. Does it look ok?

Thanks both of you for your kind support and mentorship!

• J

[blue-j]

there are some values that are a little puzzling:

DCDM XYZ : 9.334248455730954
Max RGB : 14.07103825136612
Xtreme RGB : 2.8275862068965516

But these spaces are not of much concern and can be ignored for our purposes here, just thought I'd report.

• J