Python bindings for libjpeg-turbo using pybind11. It's using Version 3 of the TurboJPEG C API.
It implements the three functions transform, compress and decompress. For more details see the function docstrings or the TurboJPEG C API docs https://rawcdn.githack.com/libjpeg-turbo/libjpeg-turbo/main/doc/html/group___turbo_j_p_e_g.html
pip install turbojpeg
Rotate JPEG file losslessly by 90 degrees.
import turbojpeg
with open("image.jpg", "rb") as fr:
img_data = fr.read()
# lossless rotation. fails if lossless rotation is not possible
# for example when the image dimensions are not multiples of 16
img_data_rot = turbojpeg.transform(img_data, op=turbojpeg.OP.ROT90)
with open("image-rotated.jpg", "xb") as fw:
fw.write(img_data_rot)Save random image as JPEG with arithmetic coding.
import numpy as np
import turbojpeg
arr = np.random.randint(0, 256, (256, 256, 3), dtype=np.uint8) # random color image
quality = 90
result = turbojpeg.compress(arr, quality, turbojpeg.SAMP.Y420, arithmetic=True)
with open("arithmetic.jpg", "xb") as fw:
fw.write(result)The Python setup.py expects the libjpeg-turbo binaries in libjpeg-turbo-build. You can either build it yourself or download precompiled binaries. For more details see https://github.com/libjpeg-turbo/libjpeg-turbo/blob/main/BUILDING.md.
GitHub actions CI builds wheels for Windows 32 and 64-bit, manylinux 64-bit, musllinux 64-bit and MacOS 64-bit. 32-bit Linux builds fail due some weird NASM error. Not all wheels are tested, since some of the test dependencies are not available for all platforms.
python -m unittest discover -s tests