Separable Four Points Fundamental Matrix
share
We present an approach for the computation of the fundamental matrix based on epipolar homography decomposition. We analyze the geometrical meaning of the decomposition-based representation and show that it guarantees a minimal number of RANSAC samples, on the condition that four correspondences are on an image line. Experiments on real-world image pairs show that our approach successfully recovers such four correspondences, provides accurate results and requires a very small number of RANSAC iterations.
READ FULL TEXT
