High-performance image processing library for Python, written in Rust with parallel processing capabilities.
img_rust provides fast image blending operations for multi-dimensional image stacks, particularly useful for microscopy and scientific imaging applications. The library leverages Rust's performance and safety guarantees while providing a seamless Python interface through PyO3 and numpy integration.
- Parallel Processing: Process multiple image frames simultaneously using Rayon
- Screen Blend Mode: Implements photoshop-style screen blending for image compositing
- 4D Array Support: Handles image stacks with dimensions (frames, height, width, channels)
- Zero-Copy Integration: Efficient data sharing between Python and Rust via numpy arrays
- Memory Safe: Leverages Rust's ownership system for safe memory management
- Python 3.7+
- Rust (latest stable)
- numpy
git clone https://github.com/wl-stepp/img_rust.git
cd img_rust
pip install maturin
maturin developimport numpy as np
import img_rust
# Create two image stacks (frames, height, width, channels)
stack1 = np.random.randint(0, 255, (10, 100, 100, 4), dtype=np.uint8)
stack2 = np.random.randint(0, 255, (10, 100, 100, 4), dtype=np.uint8)
# Apply screen blend mode across all frames in parallel
result = img_rust.screen_stack_py(stack1, stack2)The screen blending operation is parallelized across frames, providing significant speedup for large image stacks compared to pure Python implementations. Typical performance improvements of 5-20x have been observed depending on stack size and hardware.
Applies screen blend mode between two image stacks.
Parameters:
stack1(numpy.ndarray): First image stack with shape (frames, height, width, 4)stack2(numpy.ndarray): Second image stack with shape (frames, height, width, 4)
Returns:
numpy.ndarray: Blended image stack with same shape as inputs
Note: Both input arrays must have 4 channels (RGBA) and be of dtype uint8.
The screen blend mode formula for each color channel is:
result = 255 - 255 * (1 - color1/255) * (1 - color2/255)
Alpha channel blending:
alpha = 255 - ((255 - alpha1) * (255 - alpha2) / 255)
- Microscopy: Blending fluorescence channels from time-lapse imaging
- Scientific Visualization: Combining multiple data layers
- Image Processing Pipelines: High-performance batch processing
- Research Applications: Fast prototyping with performance-critical operations
Contributions are welcome! Please feel free to submit issues or pull requests.
MIT License - see LICENSE file for details.
Willi Stepp - @wl-stepp