This repository contains a collection of routines for you to compute matrices to convert between different color spaces.
Matrices from Wikipedia may not always be accurate, since they may suffer from multiple round-off errors.
The sources of the algorithms, constants, coefficients, and formulas are noted in the program code. Unless restricted by applicable low, the source code in this repository is released to the Public Domain.
Feel free to copy-paste my code into yours, or rewrite it into your favorite programming language. I would be very happy if every piece of software in the world can use the correct formula to calculate colors and be free from forgetting to perform gamma-correction.
coords.py: Defines RGB, XYZ, xyY, xy.gammas.py: Defines EOTF (gamma function) and OETF (inverse gamma function).primaries.py: Relative colorimetric conversion between RGB-based color spaces (sRGB, P3, BT.2020, ACES).whites.py: Relative colorimetric conversion between white points (aka. chromatic adaptation).
>>> from mpmath import mp
>>> mp.prec = 64 # Set precision in bits, see below
>>>
>>> from primaries import *
>>> RGB_to_RGB(sRGB, ACES2065_1)
matrix(
[['0.439632981919491366', '0.382988698151553879', '0.177378319928954755'],
['0.0897764429588423266', '0.813439428748980621', '0.0967841282921770527'],
['0.0175411703831727361', '0.111546553302387013', '0.870912276314440251']])
>>>
>>> RGB_to_RGB(DisplayP3, BT_2020)
matrix(
[['0.75383303436172182', '0.198597369052616304', '0.0475695965856618765'],
['0.0457438489653583301', '0.941777219811693547', '0.0124789312229481225'],
['-0.00121034035451832443', '0.0176017173010898941', '0.98360862305342843']])There is no support for LAB colorspaces (CIELAB, Oklab) because:
- It is almost impossible to use an incorrect formula.
- The conversion cannot be expressed with a single matrix.
- You can
import colourto perform the conversion.
There is no support for CMYK colorspaces (FOGRA39, SWOP, etc.) because:
- CMYK is too complicated. It requires a look-up table, Sakamoto's interpolation method and some complex formulas for calculating dot-gain.
- You might need to use Little CMS or Argyll CMS.
Most code in this repository relies on mpmath to perform arbitrary-precision floating point calculations. Refer to its documentation to learn how to set the precision prior to calculation.
This is because the common usage of this repository is to generate a conversion matrix, and you want to minimize round-off errors. If you need to copy-paste my code into your own software, feel free to lower the precision to 53-bit, 32-bit, or even 24-bit.
Note: Code in gammas.py uses standard 53-bit precision only, in order to be
compatible with numpy.