python-lcms2

Python binding for the LittleCMS2 color management library.

Installation

pip install python-lcms2

Pre-built wheels are available on PyPI for Windows (x86-64, ARM64), Linux (x86-64, ARM64), and macOS (Intel, Apple Silicon) on Python 3.10 and later. Windows ARM64 requires Python 3.11 or later. For other platforms, see Building from source.

Usage

import lcms2

Profiles

A Profile can be created from a built-in name, a file on disk, or raw bytes in memory:

srgb    = lcms2.Profile("sRGB")
prophoto = lcms2.Profile("ProPhoto RGB")
from_file = lcms2.Profile(filename="CMYK.icm")

with open("CMYK.icm", "rb") as f:
    from_bytes = lcms2.Profile(buffer=f.read())

Built-in profiles provided by LittleCMS: sRGB, Lab, XYZ.

Additional built-in RGB profiles: Adobe RGB (1998), Apple RGB, Best RGB, Beta RGB, Bruce RGB, CIE RGB, ColorMatch RGB, Don RGB 4, ECI RGB v2, Ekta Space PS5, NTSC RGB, PAL/SECAM RGB, ProPhoto RGB, SMPTE-C RGB, _sRGB, Wide Gamut RGB.

Transforms

A Transform converts pixel data from one profile and format to another:

lab  = lcms2.Profile("Lab")
srgb = lcms2.Profile("sRGB")

transform = lcms2.Transform(lab, "Lab_DBL", srgb, "RGB_16")

The constructor signature is:

Transform(src_profile, src_format, dst_profile, dst_format, intent="PERCEPTUAL", flags="NONE")

Format strings

Format strings follow the pattern {COLORSPACE}_{DTYPE} with an optional suffix. The dtype suffix determines the NumPy array type expected at runtime:

Suffix	NumPy dtype	Typical use
_8	uint8	8-bit images (0–255)
_16	uint16	16-bit images (0–65535)
_FLT	float32	Single-precision float
_DBL	float64	Double-precision float
_HALF_FLT	float16	Half-precision float

Common colorspace prefixes: RGB, BGR, RGBA, BGRA, GRAY, Lab, XYZ, CMYK, YCbCr, HSV, HLS.

Optional suffixes: _PLANAR (channel-first layout), _REV (ink-reversed), _SE (swap-endian), _PREMUL (premultiplied alpha).

The full list of supported format strings:

lcms2.DATA_TYPES.keys()

Rendering intents (lcms2.INTENT): PERCEPTUAL, RELATIVE_COLORIMETRIC, SATURATION, ABSOLUTE_COLORIMETRIC.

Flags (lcms2.FLAG) can be combined with commas, spaces, semicolons, or |:

transform = lcms2.Transform(lab, "Lab_DBL", srgb, "RGB_16",
                            intent="PERCEPTUAL",
                            flags="GAMUTCHECK,SOFTPROOFING")

Applying a transform

Transform.apply accepts a NumPy array or any array-compatible object and returns a transformed NumPy array. It works on single pixels, batches, and full images — the last dimension is always the channel axis:

result = transform.apply([50.0, 0.0, 0.0])   # L*a*b* neutral grey, shape (3,)

The array dtype must match the source format declared at construction time. When passing lists or other convertibles, the data is cast automatically where possible.

Converting a full image

apply treats any array shape as (..., channels), so a 2D image (H, W, C) is processed in one call:

import numpy as np
import lcms2

srgb     = lcms2.Profile("sRGB")
prophoto = lcms2.Profile("ProPhoto RGB")

transform = lcms2.Transform(srgb, "RGB_8", prophoto, "RGB_16")

image   = np.random.randint(0, 256, (480, 640, 3), dtype=np.uint8)
result  = transform.apply(image)   # shape (480, 640, 3), dtype uint16

A common use case is editing in Lab — perceptually uniform, so lightness adjustments look natural. OpenCV loads images as BGR uint8:

import cv2
import numpy as np
import lcms2
import matplotlib.pyplot as plt

image = cv2.imread("photo.jpg")               # uint8, BGR

srgb = lcms2.Profile("sRGB")
lab  = lcms2.Profile("Lab")

to_lab  = lcms2.Transform(srgb, "bgr_8",   lab,  "lab_flt")
to_srgb = lcms2.Transform(lab,  "lab_flt", srgb, "rgb_flt")

lab_image = to_lab.apply(image)
lab_image[..., 0] = np.clip(lab_image[..., 0] * 1.2, 0, 100)  # boost lightness

result = np.clip(to_srgb.apply(lab_image), 0.0, 1.0)   # float32, RGB, 0.0–1.0

plt.imshow(result)
plt.show()

Format strings are case-insensitive, so "BGR_8", "bgr_8", and "Bgr_8" are all accepted.

To save back to a file with cv2:

bgr_out = cv2.cvtColor((result * 255).astype(np.uint8), cv2.COLOR_RGB2BGR)
cv2.imwrite("output.jpg", bgr_out)

Building from source

Clone the repository with its submodules:

git clone --recurse-submodules https://github.com/wzjoriv/python-lcms2.git
cd python-lcms2

Install the build tool if needed:

pip install build --upgrade

Build a wheel:

python -m build --wheel

The wheel is written to dist/. Install it with:

pip install dist/python_lcms2*.whl

Windows path detection (Windows SDK include and library directories) is handled automatically via the registry in setup.py. If the build fails on your Windows installation, the paths may need to be adjusted manually.

For bugs, questions, and feature requests use GitHub Issues.

Acknowledgements

This project is a continuation of work by prior maintainers. The lineage of the repository is: sk1-project → RomanKosobrodov → wzjoriv.