Random Sample Consensus (RANSAC) is an important and popular robust algorithm. It is widely used in many computer vision tasks such as 3D reconstruction, wide baseline stereo, homography or epipolar geometry estimation.

One of the most popular computer vision libraries -- OpenCV -- includes RANSAC. The current existing implementation does not include recently published state-of-the-art modifications of RANSAC. Which means that OpenCV implementation could be made both faster and more accurate by integrating the recently developed RANSAC methods. Moreover, MAGSAC dispenses with the need to specify a precise inlier-outlier threshold.

Implementing recent improvements of RANSAC into OpenCV is the objective of the proposal.



Maksym Ivashechkin


  • Riba Edgar
  • Vadim Pisarevsky