🎨 Pixiq - Smart Image Compression

Intelligently compress images while preserving visual quality

Compress smarter, not harder. Let AI choose the perfect quality for your images.

📦 Installation • 🚀 Quick Start • 📚 Documentation • ❓ Why Pixiq?

---

✨ Key Features

🎯 Smart Quality Selection - Automatically finds the optimal compression quality to match your target visual quality

🖼️ Multiple Formats - Supports JPEG, PNG, WEBP, and AVIF with format-specific optimizations

📏 Intelligent Resizing - Resize images while maintaining aspect ratio and quality

🔄 Thumbnail Generation - Create smaller versions of compressed images instantly

🎨 Alpha Channel Support - Handles transparent images correctly for each format

⚡ Performance Optimized - Fast compression with efficient memory usage

🧩 Tile-Based Quality Check - PSNR analysis performed on image tiles for fast quality assessment

🛡️ Robust Validation - Comprehensive input validation and error handling

🔍 Quality Metrics - Detailed compression statistics and iteration info

---

📦 Installation

From PyPI (Recommended)

Using uv (fastest):

uv add pixiq
# or
uv pip install pixiq

Using pip:

pip install pixiq

From Source

git clone https://github.com/yourusername/pixiq.git
cd pixiq

# Using uv
uv sync --dev
uv pip install -e .

# Using pip
pip install -e .

Requirements

• Python 3.9+
• PIL (Pillow) - Image processing
• NumPy - Array operations
• pillow-avif-plugin - AVIF format support

💡 Tip: uv is significantly faster than pip and provides better dependency management.

---

🚀 Quick Start

Try it now! Run the interactive demo:

python example.py

Basic Compression

from PIL import Image
from pixiq import Pixiq

# Open your image
image = Image.open('photo.jpg')

# Compress with target quality
result = Pixiq.compress(image, perceptual_quality=0.9)

print(f"✅ Compressed! Size: {result.file_size_kb:.1f} KB, Quality: {result.selected_quality}")

Advanced Usage

# Compress with custom settings
result = Pixiq.compress(
    input=image,
    perceptual_quality=0.85,    # Target visual quality (0.0-1.0)
    max_size=2000,             # Resize if larger than 2000px
    format='WEBP',             # Force WEBP format
    hash_type='sha1',          # Use SHA1 instead of default SHA256
    output='compressed.webp'   # Save to file
)

# Access compression details
print(f"📊 Quality: {result.selected_quality}/100")
print(f"📏 Dimensions: {result.dimensions}")
print(f"💾 Size: {result.file_size_kb:.1f} KB")
print(f"🔐 Hash ({result.hash_type}): {result.hash}")
print(f"🎯 Achieved quality: {result.best_iteration['perceptual_quality']:.3f}")

Thumbnail Generation

# Create a thumbnail from compressed image
thumbnail = result.save_thumbnail(
    max_size=500,
    output='thumbnail.webp'
)

print(f"🖼️ Thumbnail: {thumbnail.dimensions}, {thumbnail.file_size_kb:.1f} KB")

---

📚 API Reference

Pixiq.compress() ⚡

The main method for intelligent image compression with automatic quality selection.

Parameters

Parameter	Type	Default	Description
input	PIL Image	required	Input image to compress
perceptual_quality	float	0.95	Target visual quality (0.0-1.0)
tolerance	float	0.005	Quality tolerance for convergence
max_quality	int	None	Maximum compression quality (1-100)
min_quality	int	None	Minimum compression quality (1-100)
max_size	int	None	Maximum dimension (resizes if larger)
max_iter	int	5	Maximum binary search iterations
format	str	None	Force format: 'JPEG', 'PNG', 'WEBP', 'AVIF'
hash_type	str	'sha256'	Hash algorithm: 'sha1', 'sha256', etc.
output	str/BytesIO	None	Output file path or buffer

Returns

CompressionResult - Object containing compressed image and metadata

Exceptions

• TypeError - Invalid parameter types
• ValueError - Parameter values out of range
• OSError - File I/O errors

---

CompressionResult 📊

Container for compression results with convenient access methods.

Core Properties

Property	Type	Description
compressed	PIL Image	The compressed image
iterations_count	int	Number of compression attempts
iterations_info	List[dict]	Detailed info for each iteration
selected_quality	int	Final compression quality (1-100)
hash	str	Hash of compressed image (SHA1/SHA256)
hash_type	str	Hash algorithm used ('sha1', 'sha256')
file_size	int	File size in bytes
fmt	str	Image format ('jpeg', 'webp', etc.)
extra_save_args	dict	Format-specific save parameters

Computed Properties

Property	Type	Description
file_size_kb	float	File size in kilobytes
dimensions	tuple[int, int]	Image dimensions (width, height)
best_iteration	dict | None	Info about best quality match

Methods

save(output) 💾

Save compressed image to file or buffer.

result.save('output.webp')
result.save(io.BytesIO())

save_thumbnail(max_size, output=None) 🖼️

Create and save a resized version of the compressed image.

thumbnail = result.save_thumbnail(max_size=500, output='thumb.webp')

Returns: New CompressionResult with resized image

---

🎨 Supported Formats

Format	Extension	Alpha Support	Optimization
JPEG	.jpg, .jpeg	❌	Progressive, optimized
PNG	.png	✅	Lossless compression
WEBP	.webp	✅	Method 6 (max compression)
AVIF	.avif	✅	Speed 6 (balanced)

💡 Tip: Pixiq automatically detects format from file extension or uses JPEG as fallback

---

🧠 How It Works

Pixiq uses a smart binary search algorithm to find the optimal compression quality:

1. 🎯 Quality Search - Binary search over quality range (1-100) to find optimal compression
2. 📊 Tile-Based PSNR Analysis - Calculate Peak Signal-to-Noise Ratio on selected image tiles (not the entire image) for fast quality assessment. The algorithm automatically detects optimal tile size based on image type (UI vs photos) and selects tiles with the most color complexity for accurate quality measurement.
3. 🧠 Perceptual Mapping - Convert PSNR to perceptual quality using empirical formula
4. 🎪 Best Match Selection - Choose quality with minimum error from target perceptual quality
5. 🔗 Efficient Hashing - Generate SHA1/SHA256 hash from compressed data without re-encoding

Algorithm Visualization

Target Quality: 0.85
┌─────────────────────────────────────┐
│ Quality 1 ────► Quality 100         │
│     ↓              ↓              ↓ │
│  PSNR: 25.3     PSNR: 35.7      PSNR: 42.1 │
│  Perceptual: 0.3 ────► 0.8 ────► 0.95 │
│     │              │              │ │
│     └──────────────┼──────────────┘ │
│                    ▼                │
│              ✅ Best Match          │
│              Quality: 67            │
│              Error: 0.0012          │
└─────────────────────────────────────┘

---

❓ Why Pixiq?

🔍 The Problem

Traditional image compression requires manual quality tuning:

# Old way - guesswork required
image.save('output.jpg', quality=85)  # Is 85 good enough?
image.save('output.jpg', quality=75)  # Too low quality?
image.save('output.jpg', quality=80)  # Still guessing...

✅ The Solution

Pixiq automatically finds the perfect quality for your needs:

# New way - specify what you want
result = Pixiq.compress(image, perceptual_quality=0.9)
# Automatically finds quality=67 for 90% perceptual quality!

🚀 Performance Benefits

Feature	Traditional	Pixiq
Quality Control	Manual guesswork	Precise target quality
File Size	Variable, unpredictable	Optimal for quality target
Time	Multiple manual attempts	Single API call
Consistency	Depends on user expertise	Consistent, reproducible
Formats	One quality per format	Optimized per format

📈 Real-World Results

Original: photo.jpg (2.3 MB, 4000x3000)
Target: 85% perceptual quality

Format    Quality    Size     Time
JPEG      78         245 KB   0.8s
WEBP      82         198 KB   0.7s
AVIF      75         156 KB   1.2s

---

🧪 Testing

Run the comprehensive test suite:

# Install development dependencies (using uv - recommended)
uv sync --dev

# Or using pip
pip install -e ".[dev]"

# Run tests
uv run pytest tests/

# Or using pytest directly
pytest tests/

# Or run manually
python -c "from tests.test_pixiq import *; test_basic_compression()"

💡 Tip: Using uv sync --dev is the fastest way to set up the development environment!

---

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Ensure all tests pass
5. Submit a pull request

---

📄 License

MIT License - see LICENSE file for details.

---

🙏 Acknowledgments

• Built with Pillow for image processing
• Uses NumPy for efficient array operations
• Powered by uv for lightning-fast package management
• Inspired by modern image optimization techniques

---

Made with ❤️ for developers who care about image quality

⭐ Star us on GitHub • 🐛 Report Issues • 💬 Join Discussions