Awesome

homography_est

Light-weight Python bindings to perform homography estimation between two images with RANSAC from point, line or point-line correspondences. Based on RansacLib and developed by Iago Suarez, Viktor Larsson, and Rémi Pautrat.

Installation

This work relies on the following dependencies:

• Eigen (3.4.0)
• Ceres Solver (2.0.0)

After installing these two dependencies, you can clone and install the homography estimation code:

git clone --recurse-submodules git@github.com:rpautrat/homography_est.git
cd homography_est
pip install -e .

Usage

Check the examples in run_hest_test.cc and run_hest_test.py to understand how to use the library. We provide the following Python functions:

• ransac_point_homography: homography estimation from a minimal set of 4 points.
• ransac_line_homography: homography estimation from a minimal set of 4 lines.
• ransac_point_line_homography: homography estimation with hybrid RANSAC from minimal sets of 4 points or 4 lines.

We also give the option to estimate the homography in case of a pure rotation between the images. In this case, only 2 features are minimal.