azusa

Auto white balance correction for digital images using professional color science.

Features

• Bradford Chromatic Adaptation: Professional-grade white balance using perceptually accurate transform
• Ensemble Illuminant Detection: Combines Gray World, Shades of Gray, and White Patch algorithms
• Smart Adaptive Correction: Per-pixel weighting based on luminance, saturation, and warmth
• Neutral Guards: Prevents overcorrection with automatic feedback loops
• Skin Tone Preservation: Intelligent protection of natural skin colors
• Real-time Processing: Instant preview with adjustable parameters
• Download Support: Save corrected images as high-quality JPEG
• Dark/Light Theme: Clean interface with system preference detection
• Client-side Processing: No uploads required - everything runs in your browser

Usage

1. Upload Image: Drag and drop or select an image file (JPEG/PNG/WebP)
2. Automatic Analysis: App detects color temperature and warmth levels (0-100 scale)
3. Adjust Parameters:
   • Strength: Correction intensity (0-100%)
   • Preserve Skin: Protects skin tones from over-correction
   • Auto-Correct: Enables tone curve and contrast enhancement
   • Keep Whites Neutral: Prevents neutral colors from shifting
   • Neutral Bias: Fine-tune the neutral point
   • Vivid: Restores color saturation (0-100%)
4. Compare Results: Interactive before/after slider
5. Download: Click button to save corrected image

How It Works

Bradford Transform

The app uses the Bradford chromatic adaptation transform, which models human visual adaptation:

1. Illuminant Estimation: Weighted ensemble of detection methods
2. Color Space Transform: RGB → XYZ → LMS cone space
3. Diagonal Scaling: Adapt between source and target illuminants
4. Perceptual Application: Per-pixel weighting based on image characteristics

Detection Methods

• Gray World: Assumes average scene color should be neutral
• Shades of Gray: Higher-order statistics with Minkowski norm (p=6)
• White Patch: Brightest pixels as reference whites
• Dynamic Weighting: More reliance on White Patch for stronger casts

Adaptive Processing

• Luminance Weighting: More correction in mid-tones, less in highlights
• Saturation Boost: Restores colors lost during white balance
• Warmth Targeting: Focuses correction on warm/yellow areas
• Neutral Protection: Iterative feedback to prevent blue shifts

Development

# Install dependencies
pnpm install

# Start development server
pnpm run dev

# Build for production
pnpm run build

# Preview production build
pnpm run preview

Deployment

This app is configured for GitHub Pages deployment. Push to the main branch to automatically deploy via GitHub Actions.

Technical Details

• Framework: Svelte 5 with TypeScript
• Build Tool: Vite
• Performance: 12-30ms analysis on 12MP images, 120-400ms full correction
• Color Science: D65 illuminant, accurate sRGB ↔ CIELAB conversions
• Browser Compatibility: Modern browsers with Canvas API support

License

This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License.

You are free to:

• Share and redistribute the material
• Adapt, remix, and build upon the material

Under the following terms:

• Attribution — You must give appropriate credit
• NonCommercial — You may not use the material for commercial purposes

See the LICENSE file for full details.