@panda-ai/imagequant

High-performance WebAssembly port of imagequant (libimagequant) for color quantization in browsers. Achieves 2-4x faster processing compared to JS implementations.

Features

• 🔥 WASM accelerated color quantization
• 🎨 Alpha transparency support
• 📦 80kb gzipped WASM bundle
• 🚀 Zero-copy memory operations
• 📱 Web Worker compatible

Installation

npm

bun install @panda-ai/imagequant

Quick Start

  import WasmImageQuant, { PNGEncoder } from '@panda-ai/imagequant'

  const iq = await WasmImageQuant.create();
  const filePath = path.resolve(__dirname, './boat.png')
  const file = Bun.file(filePath);
  const arrayBuffer = await file.arrayBuffer();
  const imageData = await readBufferToImageData(arrayBuffer);

  const { palette, indices } = iq.quantize(imageData);

    // Generate PNG
  const pngData = PNGEncoder.encode(
    palette,
    indices,
    imageData.width,
    imageData.height,
  );

API

quantize(pixels: Uint8Array, width: number, height: number, options): QuantResult

| Option | Default | Description | |-------------|---------|---------------------------------| | maxColors | 256 | Maximum palette size (2-256) | | speed | 6 | Speed/quality tradeoff (1-10) |

Returns:

interface QuantResult {
  palette: Uint8Array  // RGBA palette (maxColors × 4 bytes)
  indices: Uint8Array  // Pixel indices (width × height bytes)
}

Performance Benchmarks

| Image Size | Colors | WASM Time | JS Time | |------------|--------|-----------|---------| | 512x512 | 128 | 68ms | 220ms | | 1024x768 | 256 | 145ms | 520ms | | 4096x2160 | 256 | 890ms | 3200ms |

Tested on Chrome 118, M1 MacBook Pro

Browser Support

| Browser | Minimum Version | |---------------|-----------------| | Chrome | 57+ | | Firefox | 53+ | | Safari | 14.1+ | | Edge | 79+ |

Advanced Usage

Web Worker Processing

TBD

Canvas Integration

TBD

Limitations

TBD

Building from Source

1. Install Rust toolchain
2. Install wasm-pack

curl https://rustwasm.github.io/wasm-pack/installer/init.sh -sSf | sh

3. Build package

wasm-pack build --target web

License

MIT © Panda Intelligence. Commercial licenses available.