Apple HDR Gain Map Reconstruction

This Jupyter notebook reconstructs HDR images from Apple's SDR HEIC images and their associated gain maps.

Quick Start on MyBinder

1. Go to https://mybinder.org/
2. Enter the GitHub repository URL
3. Click "launch"

Technical Details

The HDR Reconstruction Formula

The notebook implements Apple's gain map formula:

L_hdr = L_sdr × (1 + (Headroom - 1) × GainMap)

Where:

• L_sdr is the SDR (Standard Dynamic Range) image
• GainMap is the additional brightness information
• Headroom is extracted from XMP metadata (how much brighter HDR is vs SDR)

Color Space Pipeline

1. Input: Display P3 color space (Apple devices)
2. Linearization: Remove gamma encoding
3. Gain Application: Apply the formula above
4. Color Adaptation: Convert to BT.2020 using Bradford chromatic adaptation
5. PQ Encoding: Apply Perceptual Quantizer transfer function
6. Output: 16-bit PNG with cICP metadata chunk

Output Specifications

• Format: 16-bit PNG
• Color Primaries: BT.2020 (ITU-R Recommendation BT.2020)
• Transfer Function: PQ/ST.2084 (Perceptual Quantizer)
• Reference White: 203 nits
• Peak Brightness: 1000 nits
• Metadata: cICP chunk for HDR signaling

Troubleshooting

"Files not found" error

• Ensure you've created the data/ folder
• Verify your files are named exactly as expected
• Check file extensions are correct (.HEIC, .png)

Import errors

• On mybinder, dependencies should install automatically
• If running locally, use: pip install -r requirements.txt

"No XMP metadata" warning

• The gain map PNG should contain XMP metadata with headroom value
• If missing, default headroom of 1.0 is used (no HDR enhancement)

License

This notebook is provided for educational purposes.