coloria

Tools for color research.

Installation

Install Coloria from PyPI with

pip install coloria

To run Coloria, you need a license. See here for more info.

Illuminants, observers, white points

Illuminants	CIE 1931 Observer

import coloria
import matplotlib.pyplot as plt

illu = coloria.illuminants.d65()
plt.plot(illu.lmbda_nm, illu.data)
plt.xlabel("wavelength [nm]")
plt.show()

The following illuminants are provided:

• Illuminant A ("indoor light", coloria.illuminants.a(resolution_in_nm))
• Illuminant C (obsolete, "North sky daylight", coloria.illuminants.c())
• Illuminants D ("natural daylight", coloria.illuminants.d(nominal_temp) or coloria.illuminants.d65() etc.)
• Illuminant E (equal energy, coloria.illuminants.e())
• Illuminant series F ("fluorescent lighting", coloria.illuminants.f2() etc.)

Observers:

• CIE 1931 Standard 2-degree observer (coloria.observers.coloria.observers.cie_1931_2())
• CIE 1964 Standard 10-degree observer (coloria.observers.coloria.observers.cie_1964_10())

Color appearance models

Color appearance models (CAMs) predicts all kinds of parameters in color perception, e.g., lightness, brightness, chroma, colorfulness, saturation etc. Since these values depend on various factors, such as the surrouning, the models are initialized with various different parameters.

CAMs can be used to construct color spaces (see below).

The color appearance models available in coloria are

• CIECAM02 / CAM02-UCS

  import coloria
  
  ciecam02 = coloria.cam.CIECAM02(
      surround_type="average",  # or dim, dark
      background_luminance_percent=20,
      adapting_luminance_cd_m2=100,
  )
  
  xyz = [19.31, 23.93, 10.14]
  corr = ciecam02.from_xyz100(xyz)
  # then work with those values:
  corr.lightness
  corr.brightness
  corr.chroma
  corr.hue_composition
  corr.hue_angle_degrees
  corr.colorfulness
  corr.saturation
  

• CAM16 / CAM16-UCS

  import coloria
  
  cam16 = coloria.cam.CAM16("average", 20, 100)
  

• ZCAM

  import coloria
  
  cam16 = coloria.cam.ZCAM("average", 20, 100, 20)
  

Color coordinates and spaces

Color coordinates are handled as NumPy arrays or as ColorCoordinates, a thin wrapper around the data that retains the color space information and has some handy helper methods. Color spaces can be instantiated from the classes in coloria.cs, e.g.,

import coloria

coloria.cs.CIELAB()

Most methods that accept such a colorspace also accept a string, e.g., cielab.

As an example, to interpolate two sRGB colors in OKLAB, and return the sRGB:

from coloria.cs import ColorCoordinates

# you can also plug in large numpy arrays instead of two lists here
c0 = ColorCoordinates([1.0, 1.0, 0.0], "srgb1")  # yellow
c1 = ColorCoordinates([0.0, 0.0, 1.0], "srgb1")  # blue

# naive interpolation gives [0.5, 0.5, 0.5], a mid gray

# convert to OKLAB
c0.convert("oklab")
c1.convert("oklab")

# interpolate
c2 = (c0 + c1) * 0.5

c2.convert("srgbhex", mode="clip")

print(c2.color_space)
print(c2.data)

#6cabc7

All color spaces implement the two methods

vals = colorspace.from_xyz100(xyz)
xyz = colorspace.to_xyz100(vals)

for conversion from and to XYZ100. Adding new color spaces is as easy as writing a class that provides those two methods. The following color spaces are already implemented:

• XYZ (coloria.cs.XYZ(100), the parameter determining the scaling)

• xyY (coloria.cs.XYY(100), the parameter determining the scaling of Y)

• sRGB (coloria.cs.SRGBlinear(), coloria.cs.SRGB1(), coloria.cs.SRGB255(), coloria.cs.SRGBhex())

• HSL and HSV (coloria.cs.HSL(), coloria.cs.HSV()) These classes also have the two methods

  from_srgb1()
  to_srgb1()
  

  for direct conversion from and to standard RGB.

• OSA-UCS (coloria.cs.OsaUcs()), 1947

• CIE 1960 UCS (coloria.cs.CIE1960UCS()), 1960

• CIEUVW (coloria.cs.CIEUVW()), 1964

• CIELAB (coloria.cs.CIELAB()), 1976

• CIELUV (coloria.cs.CIELUV()), 1976

• RLAB (coloria.cs.RLAB()), 1993

• IPT (coloria.cs.IPT()), 1998

• DIN99 and its variants DIN99{b,c,d} (coloria.cs.DIN99()), 1999

• CAM02-UCS, 2002

  import coloria
  
  cam02 = coloria.cs.CAM02("UCS", "average", 20, 100)
  

  The implementation contains a few improvements over the CIECAM02 specification (see here).

• CAM16-UCS, 2016

  import coloria
  
  cam16ucs = coloria.cs.CAM16UCS("average", 20, 100)
  

  The implementation contains a few improvements over the CAM16 specification (see here).

• SRLAB2 (coloria.cs.SRLAB2())

• J_za_zb_z (coloria.cs.JzAzBz()), 2017

• ICtCp (coloria.cs.ICtCp()), 2018

• I_GP_GT_G (coloria.cs.IGPGTG()), 2020

• proLab (coloria.cs.PROLAB()), 2020

• Oklab (coloria.cs.OKLAB()), 2020

• OkLCh (coloria.cs.OKLCH()), 2020

• HCT (coloria.cs.HCT()/ HCTLAB (coloria.cs.HCTLAB()), 2022

All methods in coloria are fully vectorized, i.e., computation is really fast.

Color difference formulas

coloria implements the following color difference formulas:

• CIE76

  coloria.diff.cie76(lab1, lab2)
  

• CIE94

  coloria.diff.cie94(lab1, lab2)
  

• CIEDE2000

  coloria.diff.ciede2000(lab1, lab2)
  

• CMC l:c

  coloria.diff.cmc(lab1, lab2)
  

Chromatic adaptation transforms

coloria implements the following CATs:

• von Kries

  cat, cat_inv = coloria.cat.von_kries(whitepoint_source, whitepoint_destination)
  xyz1 = cat @ xyz0
  

• Bradford (coloria.cat.bradford)

• sharp (coloria.cat.sharp)

• CMCCAT2000 (coloria.cat.cmccat2000)

• CAT02 (coloria.cat.cat02)

• CAT16 (coloria.cat.cat16)

• Bianco-Schettini (coloria.cat.bianco_schettini)

Gamut visualization

coloria provides a number of useful tools for analyzing and visualizing color spaces.

sRGB gamut

CIELAB	CAM16-UCS	Oklab

The sRGB gamut is a perfect cube in sRGB space, and takes curious shapes when translated into other color spaces. The above images show the sRGB gamut in different color spaces.

import coloria

p = coloria.plot_rgb_gamut(
    "cielab",  # or coloria.cs.CIELAB()
    n=51,
    show_grid=True,
)
p.show()

For more visualization options, you can store the sRGB data in a file

import coloria

coloria.save_rgb_gamut("srgb.vtk", "cielab", n=51)
# all formats supported by https://github.com/coloria-dev/meshio

and open it with a tool of your choice. See here for how to open the file in ParaView.

For lightness slices of the sRGB gamut, use

import coloria

p = coloria.plot_rgb_slice("cielab", lightness=50.0, n=51)
p.show()
# or
# p.screenshot("screenshot.png")

Surface color gamut

XYZ	CIELAB	CAM16-UCS

Same as above, but with the surface color gamut visible under a given illuminant.

import coloria

illuminant = coloria.illuminants.d65()
observer = coloria.observers.cie_1931_2()

p = coloria.plot_surface_gamut(
    "xyz100",  # or coloria.cs.XYZ(100)
    observer,
    illuminant,
)
p.show()

The gamut is shown in grey since sRGB screens are not able to display the colors anyway.

The visible gamut

xyY	JzAzBz	Oklab

Same as above, but with the gamut of visible colors up to a given lightness Y.

import coloria

observer = coloria.observers.cie_1931_2()

colorspace = coloria.cs.XYZ(100)

p = coloria.plot_visible_gamut(colorspace, observer, max_Y1=1)
p.show()

The gamut is shown in grey since sRGB screens are not able to display the colors anyway.

For slices, use

import coloria

plt = coloria.plot_visible_slice("cielab", lightness=0.5)
plt.show()

Color gradients

With coloria, you can easily visualize the basic color gradients of any color space. This may make defects in color spaces obvious, e.g., the well-known blue-distortion of CIELAB and related spaces. (Compare with the hue linearity data below.)

import coloria

plt = coloria.plot_primary_srgb_gradients("cielab")
plt.show()

CIELAB	DIN99	OKLAB

Experimental data

coloria contains lots of experimental data sets some of which can be used to assess certain properties of color spaces. Most data sets can also be visualized.

Color differences

xyY	CIELAB	CAM16

Color difference data from MacAdam (1974). The above plots show the 43 color pairs that are of comparable lightness. The data is matched perfectly if the facing line stubs meet in one point.

import coloria

data = coloria.data.MacAdam1974()

cs = coloria.cs.CIELAB

plt = data.plot(cs)
plt.show()
print(coloria.data.MacAdam1974().stress(cs))

24.54774029343344

The same is available for

coloria.data.BfdP()
coloria.data.Leeds()
coloria.data.RitDupont()
coloria.data.Witt()

coloria.data.COMBVD()  # a weighted combination of the above

Munsell

xyY	CIELAB	CAM16

Munsell color data is visualized with

import coloria

cs = coloria.cs.CIELUV
plt = coloria.data.Munsell().plot(cs, V=5)
plt.show()

To retrieve the Munsell data in xyY format, use

import coloria

munsell = coloria.data.Munsell()

# munsell.h
# munsell.V
# munsell.C
# munsell.xyy

Ellipses

MacAdam ellipses (1942)

xyY (at Y=0.4)	CIELAB (at L=50)	CAM16 (at L=50)

The famous MacAdam ellipses (from this article) can be plotted with

import coloria

cs = coloria.cs.CIELUV
plt = coloria.data.MacAdam1942(50.0).plot(cs)
plt.show()

The better the colorspace matches the data, the closer the ellipses are to circles of the same size.

Luo-Rigg ellipses

xyY	CIELAB	CAM16

Likewise for Luo-Rigg.

import coloria

# xyy = coloria.cs.XYY(100)
# coloria.data.LuoRigg(8).show(xyy, 0.4)
# coloria.data.LuoRigg(8).savefig("luo-rigg-xyy.png", xyy, 0.4)

cieluv = coloria.cs.CIELUV()
plt = coloria.data.LuoRigg(8).plot(cieluv, 50)
plt.show()

Hue linearity

Ebner-Fairchild

xyY	CIELAB	CAM16

For example

import coloria

colorspace = coloria.cs.JzAzBz
plt = coloria.data.EbnerFairchild().plot(colorspace)
plt.show()

shows constant-hue data from the Ebner-Fairchild experiments in the hue-plane of some color spaces. (Ideally, all colors in one set sit on a line.)

Hung-Berns

Likewise for Hung-Berns:

xyY	CIELAB	CAM16

Note the dark blue distortion in CIELAB and CAM16.

import coloria

colorspace = coloria.cs.JzAzBz
plt = coloria.data.HungBerns().plot(colorspace)
plt.show()

Xiao et al.

Likewise for Xiao et al.:

xyY	CIELAB	CAM16

import coloria

colorspace = coloria.cs.CIELAB
plt = coloria.data.Xiao().plot(colorspace)
plt.show()

Lightness

Fairchild-Chen

xyY	CIELAB	CAM16

Lightness experiment by Fairchild-Chen.

import coloria

cs = coloria.cs.CIELAB
plt = coloria.data.FairchildChen("SL2").plot(cs)
plt.show()

Articles

• Algorithmic improvements for the CIECAM02 and CAM16 color appearance models, Nico Schlömer, 2018
• On the conversion from OSA-UCS to CIEXYZ, Nico Schlömer, 2019