PyJPEG

A pure Python encoder and decoder for the JPEG image format, aiming to be a complete implementation of all the JPEG standards still in active use — including processes that most libraries skip, such as arithmetic coding, lossless coding, and JPEG-LS.

Being pure Python (no C extensions), it is not intended to be the fastest or most efficient way to decode/encode JPEGs in Python — libraries like pylibjpeg or Pillow will perform much better for general image loading/saving. PyJPEG is instead most useful when you want to understand how JPEG works, or need to analyze or generate JPEG files at the bitstream/segment level in plain, readable Python.

Features

• Pure Python — no compiled dependencies, easy to install anywhere Python runs
• Simple, inspectable object model (Image, Component, FileReader, FileWriter)
• Work at whichever level suits your task — manipulate files at the raw stream/segment level, or through high-level Image objects
• Comprehensive JPEG feature support, including modes most libraries omit

Supported JPEG coding processes:

• Baseline sequential DCT — the standard JPEG process used by most .jpg files
• Progressive DCT — encodes images in multiple passes of increasing detail, including successive approximation scans, so images can be displayed at low quality before the full file has loaded
• Lossless coding — a predictive coding mode that reconstructs the exact original pixel values, with no loss of image quality
• 12-bit precision — supports 12-bit sample precision as well as the standard 8-bit, which many other JPEG libraries don't support or make difficult to use
• Arithmetic coding — an alternative entropy coder to Huffman, supported across all modes
• JPEG-LS — the separate lossless/near-lossless standard (ITU-T T.87), including LSE (JPEG-LS preset parameters) support
• Huffman table optimization — generate optimized Huffman tables rather than using the standard/default ones
• Obscure marker support — including rarely-implemented markers like DNL (define number of lines), which most libraries skip

Also reads and writes the metadata formats JPEG files are commonly wrapped in, including JFIF (with thumbnails), Exif, SPIFF, and the Adobe APP14 marker used to signal color space (RGB/CMYK, YCbCr, or YCCK).

Installation

Install from PyPI:

pip install pyjpeg

Usage

See API documentation for details.

Reading a JPEG

import pyjpeg

reader = pyjpeg.FileReader(open('test.jpg', 'rb'))
image = pyjpeg.Image.read(reader)
print(image.components[0].samples)

Writing a JPEG

import pyjpeg

# A tiny 8x8 single-component (greyscale) image
samples = [
    0,   0,   0,   0,   0,   0,   0,   0,
    0,   0,   0,   255, 255, 255, 0,   0,
    0,   0,   0,   0,   255, 0,   0,   0,
    0,   0,   0,   0,   255, 0,   0,   0,
    0,   0,   255, 0,   255, 0,   0,   0,
    0,   0,   255, 0,   255, 0,   0,   0,
    0,   0,   0,   255, 0,   0,   0,   0,
    0,   0,   0,   0,   0,   0,   0,   0,
]
out_image = pyjpeg.Image(8, 8, [pyjpeg.Component(1, samples)])
writer = pyjpeg.FileWriter(open('test_out.jpg', 'wb'))
out_image.write(writer)

Contributing

PyJPEG is an open source project with an LGPL-3.0 license. It is still under active development, and APIs may change between releases. Contributions, bug reports, and testing against real-world JPEG files are welcome — see the project homepage to get started.

Alternatives

If you just need fast, general-purpose JPEG reading/writing in Python (not a pure-Python implementation, and not full standards coverage), consider:

• Pillow — the standard general-purpose Python imaging library, backed by libjpeg
• pylibjpeg — a Python framework for JPEG decoding built on compiled plugins, with a focus on DICOM use cases