FIM Examples

This directory contains examples demonstrating the FIM (FastImage) library's capabilities.

Python Examples

simple_resize_example.py

A minimal example showing the complete workflow requested:

1. Load image with fastslide
2. Print metadata
3. Crop the center
4. Rescale with MagicKernel2021
5. Convert to PIL Image

Usage:

python simple_resize_example.py slide.svs

Output:

• Prints slide dimensions and metadata
• Saves output.png with the processed result

python_resize_example.py

A comprehensive example demonstrating:

• Loading whole slide images with fastslide integration
• Metadata extraction and display
• Center cropping
• High-quality resizing with different kernels
• Subpixel-accurate box parameter usage
• Kernel comparison (Lanczos2 vs Lanczos3 vs Magic2021)
• Converting to PIL Image for further processing

Usage:

python python_resize_example.py slide.svs 512
python python_resize_example.py slide.mrxs 1024

Outputs:

• output_512x512.png - Main result
• output_with_box.png - Subpixel box example
• output_lanczos2_(fastest).png - Lanczos2 kernel result
• output_lanczos3_(default).png - Lanczos3 kernel result
• output_magic2021_(highest_quality).png - Magic2021 kernel result

Key Concepts

Lazy Evaluation

All FIM operations are lazy - they build a computation graph but don't execute until you:

• Call .to_numpy() to convert to NumPy array
• Call .write_png() or .write_tiff() to save to file

This allows efficient processing of large images by only computing what's needed.

Resampling Kernels

• LANCZOS2: Fastest, radius 2, good quality
• LANCZOS3: Default, radius 3, excellent quality/speed balance
• MAGIC2021: Highest quality, radius 4.5, minimal ringing artifacts

Box Parameter

The Box class allows specifying source regions with floating-point pixel coordinates:

box = fim.Box(x1=10.5, y1=20.5, x2=510.5, y2=520.5)

This is useful for:

• Subpixel shifts (like MRXS tile alignment)
• Precise region extraction combined with scaling
• Avoiding rounding errors in coordinate transformations

Requirements

pip install numpy pillow

Building and Running

If you've built FIM with Bazel, the Python bindings are available in the workspace:

# Build the Python bindings
bazelisk build //aifo/fimage/python:fim

# Run tests to verify installation
bazelisk test //aifo/fimage/src/python:resize_test

# Run the example
python3 examples/simple_resize_example.py /path/to/slide.svs