scikit-surgeryopencvcpp

Image-guided surgery functions, in C++, using OpenCV, wrapped in Python.

Part of the SNAPPY software project, developed at the Wellcome EPSRC Centre for Interventional and Surgical Sciences, University College London (UCL).

Features

• Stereo triangulation (Hartley/Zisserman SVD method and midpoint-of-shortest-distance)
• Calibration dot/fiducial detection
• Point masking utilities

Installing

pip install scikit-surgeryopencvcpp

Building from source

Requires CMake 3.15+ and a C++17 compiler. All dependencies (OpenCV, pybind11) are fetched automatically via CMake FetchContent.

Build a wheel (for pip install)

pip install build
python -m build --wheel

This produces a wheel in dist/ that you can install with:

pip install dist/scikit_surgeryopencvcpp-*.whl

For a faster development cycle (reuses your existing build artifacts):

pip wheel . -v --no-build-isolation

C++ development (no Python)

cmake -B build -DBUILD_TESTING=ON -DBUILD_PYTHON_BINDINGS=OFF -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release --parallel
ctest --test-dir build --build-config Release --output-on-failure

Full build with Python bindings

Requires NumPy and pybind11 in your environment:

pip install numpy pybind11
cmake -B build -DBUILD_TESTING=ON -DBUILD_PYTHON_BINDINGS=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release --parallel
ctest --test-dir build --build-config Release --output-on-failure

Python API

import numpy as np
import sksurgeryopencvpython as sks

# Triangulate stereo point pairs
points_3d = sks.triangulate_points_using_hartley(
    image_points,       # Nx4 array (left_x, left_y, right_x, right_y)
    left_intrinsics,    # 3x3 camera matrix
    right_intrinsics,   # 3x3 camera matrix
    rotation_matrix,    # 3x3 left-to-right rotation
    translation_vector  # 3x1 left-to-right translation
)

Contributing

1. Raise an issue on GitHub
2. Fork the repository
3. Create a feature branch: <issue-number>-<short-description>
4. Submit a pull request

License

Copyright 2018 University College London. Released under the BSD-3 license.

Acknowledgements

Supported by Wellcome and the EPSRC.