High-performance image tools for Python

mu-image is a cross-platform toolkit for software engineers who process 2D images in Python. The available set of image-processing primitives are highly-optimized and can be chained to run sequentially in native code. Where equivalent functions exist mu-image should be much faster and use less memory than numpy or opencv. But interfacing with both those frameworks is seamless.

Included in the package is also a fully spec compliant DNG RAW decoder and encoder that is built completely on mu-image.

The repository contains two packages:

• muraw: A Python library and CLI for DNG I/O, rendering, metadata, and multi-threaded batch processing. Import muraw. Install with pip install muraw.
• mu-dng-converter: A desktop GUI on muraw and PyWebView for batch convert and transcode.

---

Primary use cases

• High performance image processing: basic imaging primitives to develop your own pipelines.
• DNG in your app: read, write, and render DNG.
• DNG transcode: Re-encode a DNG with a different codec (uncompressed, JPEG XL lossless/lossy) and/or update TIFF/DNG metadata, without a full develop.
• Batch convert: RAW/DNG sequences to TIFF (8- or 16-bit), JPEG, or JPEG XL, including multi-core folder jobs.
• DNG to video: Stitch a folder of DNG frames into MP4/H.264.
• FITS to DNG: Convert scientific .fits to DNG for Photoshop, Lightroom, or Camera Raw.

---

Key technical features

RAW rendering (muraw)

• Demosaicing: BILINEAR, VNG, RCD, EA, EA_FAST, and OPENCV_EA.
• Opcode and color: Linearization, DNG opcodes, color matrices, and tone curves.
• macOS: Built-in renderer or Apple Core Image.
• XMP: Temperature, tint, exposure, curves, and radial distortion when present.
• CLI: Render, transcode, and inspect DNG files.

Desktop app (mu-dng-converter)

A batch-scale GUI: select files or a folder of DNG (or FITS). No Python install if you use the pre-built binary.

Render develops to TIFF, JPEG, JPEG XL, or MP4. Transcode stays in DNG: change the codec and/or rewrite metadata without a full develop.

• Transcode: Uncompressed or JPEG XL (lossless/lossy); set or strip tags, shift timestamps, timezone.
• Render control: White balance (presets or Kelvin/tint), exposure, output bit depth, resolution scale (0.125×–1.0×) for previews.
• Video: Codec, resolution, frame rate, and CRF for MP4 from raw frames.
• Convert FITS to DNG: Histogram auto-exposure so files open at a usable baseline; AVM XMP from WCS and instrument headers; JPEG previews and pyramids for fast browse.

---

Supported platforms

• Windows (x86_64)
• macOS (Intel and Apple Silicon)
• Linux (Ubuntu, Debian, Fedora)
• Raspberry Pi (ARM64)

---

Getting started

Desktop app (macOS, Windows, Linux): Download a pre-built binary from the mu-dng-converter releases page. No Python install required. See mu-dng-converter/README.md.

Python library: See muraw/README.md for install, API, and CLI.

Release history: See CHANGELOG.md.