up previous
Up: Computer Vision IT412 Previous: Image rectification

Computing the epipolar lines

 Given a point m that has coordinates (u1,v1) in the first retinal plane $\Pi$,it is known that its correspondence and the fundamental matrix are related by:

\begin{displaymath}
\left(u \quad v \quad 1 \right) \, \mbox{\bf F}
 \left(\begin{array}
{c} u_1 \\  v_1\\  1 \end{array}\right) = 0.\end{displaymath}

This equation can be expanded to

\begin{displaymath}
(F_{11} u_1 + F_{12} v_1 + F_{13}) \, u +
 (F_{21} u_1 + F_{22} v_1 + F_{23}) \, v +
 (F_{31} u_1 + F_{32} v_1 + F_{33}) = 0,\end{displaymath}

where u1 and v1 are known entities, u and v are variables. This is the equation of a line in the second retinal plane. It is the epipolar line on which m' must lie.

Given a point m that has coordinates (u1,v1) in the first retinal plane $\Pi$,to compute the epipolar line l$_{\bf m}$ that contains m, we must first compute the epipole e $\in \Pi$.

e $\equiv$ intersection point of all epipolar lines in $\Pi$.

The coordinates of l$_{\bf m}$ can then be computed as the cross-product of e and m.



Robyn Owens
10/29/1997