This paper studies error characterization of the factorization method for 3
-D shape and motion recovery from image sequences using matrix perturbation
theory and covariance propagation for linear models. Given the 2-D project
ions of a set of feature points across multiple image frames and small pert
urbations/covariances of the feature point coordinates, first-order perturb
ation and covariance matrices of 3-D shape and motion are derived and valid
ated with the ground truth. This work provides quantitative analysis of err
or sensitivity of 3-D shape and motion estimation subject to small feature
correspondence errors. It can be used to pinpoint system performance, such
as which point/frame/coordinate has relatively higher uncertainty, and gain
insight for further improvement. We show the 3-D shape uncertainty as elli
psoids on top of the 3-D reconstruction as an enhanced visualization, leadi
ng to better use of the factorization method in engineering applications. E
xperimental results are demonstrated to support the analysis and to show ho
w the error analysis and error measures can be used. (C) 2001 Academic Pres
s.