A new method is presented for robustly estimating multiple view relations f
rom point correspondences. The method comprises two parts. The first is a n
ew robust estimator MLESAC which is a generalization of the RANSAC estimato
r. It adopts the same sampling strategy as RANSAC to generate putative solu
tions, but chooses the solution that maximizes the likelihood rather than j
ust the number of inliers. The second part of the algorithm is a general pu
rpose method for automatically parameterizing these relations, using the ou
tput of MLESAC. A difficulty with multiview image relations is that there a
re often nonlinear constraints between the parameters, making optimization
a difficult task. The parameterization method overcomes the difficulty of n
onlinear constraints and conducts a constrained optimization. The method is
general and its use is illustrated for the estimation of fundamental matri
ces, image-image homographies, and quadratic transformations. Results are g
iven for both synthetic and real images. It is demonstrated that the method
gives results equal or superior to those of previous approaches, (C) 2000
Academic Press.