numpy2dng 0.1.0

Pure Python utility for creating DNG files from numpy arrays.

numpy2dng

Pure Python utility for creating Adobe DNG files from RAW image data stored in NumPy arrays.

This project is a refactor of the original PiDNG library, focused on a smaller, cleaner core and modern packaging.

Requirements

• Python 3.10+
• NumPy 2.0+

Features

• Pure Python (no C extensions)
• Uncompressed DNG output
• Supports 8/10/12/14/16-bit integer RAW (via packing for 10/12/14-bit)
• Supports float32 RAW (writes DNG 1.4-compatible files)
• “Safe packing” for 10/12/14-bit data with any image width (no special alignment required)

Non-goals / Limitations

• Not a full RAW pipeline (no demosaic, no color processing)
• Does not auto-detect camera metadata; you must provide the required DNG/TIFF tags
• No JPEG/LJ92 compression (uncompressed only)

Installation

Using uv (recommended):

uv add numpy2dng

Using pip:

pip install numpy2dng

Usage

Typical flow:

1. Prepare a 2D NumPy array of shape (height, width) with dtype uint16 (or float32).
2. Build a numpy2dng.dng.DNGTags instance and set at least:
   • Tag.ImageWidth, Tag.ImageLength, Tag.BitsPerSample
3. Convert and write a .dng using numpy2dng.core.RAW2DNG.

For performance-sensitive workflows, RAW2DNG.convert(..., file=...) can write directly to an open binary file handle (avoids an extra copy).

API Overview

• numpy2dng.core.RAW2DNG: main converter for NumPy arrays → DNG
• numpy2dng.dng.DNGTags / numpy2dng.dng.Tag: tag container and tag definitions

Development

uv sync
uv run pytest
uv run ruff check .

Credits

Based on (and inspired by) PiDNG

Future work

• Add more metadata tags and convenience methods for common camera models
• Support for additional data types and compression methods
• Improve documentation and add usage examples

License

Same as PiDNG: MIT License. See LICENSE file for details.