The optical flow observed by a moving camera satisfies, in the absence of n
oise, a special equation analogous to the epipolar constraint arising in st
ereo vision. Computing the "flow fundamental matrix" of this equation is an
essential prerequisite to undertaking three-dimensional analysis of the fl
ow. This article presents an optimal formulation of the problem of estimati
ng this matrix under an assumed noise model. This model admits independent
Gaussian noise that is not necessarily isotropic or homogeneous. A theoreti
cal bound is derived for the accuracy of the estimate. An algorithm is then
devised that employs a technique called renormalization to deliver an esti
mate and then corrects the estimate so as to satisfy a particular decomposa
bility condition. The algorithm also provides an evaluation of the reliabil
ity of the estimate. Epipoles and their associated reliabilities are comput
ed in both simulated and real-image experiments. Experiments indicate that
the algorithm delivers results in the vicinity of the theoretical accuracy
bound. (C) 2000 SPIE and IS&T. [S1017-9909(00)01202-2].