Review of Fundamental Matrix Estimation Methods

Abstract

This paper reviews the estimation of Fundamental Matrix from a set of matched point correspondences between two views of an image taken by stereo camera using numerical computation. Fundamental Matrix is a 3 by 3 matrix which relates the points projected in one image plane to the corresponding points in another image plane. The Geometry which relates the corresponding image point is called as epipolar geometry, which is an intrinsic projective geometry that is widely used in two view (stereoscopic) and multi-view geometry. There exists a vast number of methods to estimate the fundamental matrix which are categorized as linear methods, repetitive methods and robust methods. This problem has many similarities with the problem of estimating a conic (conic fitting) from a set of two dimensional points {xi,yi} or a quadratic from the set of three dimensional points.