An algorithm for estimating optical flow based on the discrete-time rigid motion of an uncalibrated camera through a static planar world is presented. The algorithm is based on differential measurements, and estimates are computed within a multi-scale decomposition and propagated from coarse to fine scales. The algorithm is designed to operate properly with large displacements (i.e., large velocities or low frame rates). The quality of the algorithm is demonstrated by comparing with a simple (smoothness-based) multi-scale optical flow estimator and an instantaneous-time direct method.