Python envelope for the popular C library libjpeg for handling JPEG files.

libjpeg offers full control over compression and decompression and exposes DCT coefficients and quantization tables.

Simply install the package with pip3

pip install jpeglib

⚠️ This will install jpeglib together with every integrated version of libjpeg, libjpeg-turbo and mozjpeg. It takes longer to install than the package.

Import the library in Python 3

import jpeglib

Get discrete cosine transform (DCT) coefficients and quantization matrices as numpy array

im = jpeglib.read_dct("input.jpeg")
im.Y; im.Cb; im.Cr; im.qt

You get luminance DCT, chrominance DCT and quantization tables.

Write the DCT coefficients back to a file with

im.write_dct("output.jpeg")

Decompress the input.jpeg into spatial representation in numpy array with

im = jpeglib.read_spatial("input.jpeg")
im.spatial

You can specify parameters such as output color space, DCT method, dithering, etc.

Write spatial representation in numpy arrray back to file with

im.write_spatial("output.jpeg")

You can specify input color space, DCT method, sampling factor, output quality, smoothing factor etc.

You can find all the details in the documentation.

It is possible to choose, which version of libjpeg should be used.

jpeglib.version.set('6b')

Currently jpeglib supports all versions of libjpeg from 6b to 9e, libjpeg-turbo 2.1.0 and mozjpeg 4.0.3. Their source codes is baked inside the package and thus distributed with it, avoiding external dependency.

Get currently used libjpeg version by

version = jpeglib.version.get()

You can also set a libjpeg version for a scope only.

jpeglib.version.set('6b')
im = jpeglib.read_spatial('image.jpeg') # using 6b
with jpeglib.version('9e'):
    im = jpeglib.read_spatial('image.jpeg') # using 9e
im = jpeglib.read_spatial('image.jpeg') # using 6b again

Developed by Martin Benes.