We present an iterative algorithm for the recovery of 2-D motion, i.e., an algorithm for the determination of a transformation which maps one image onto another. The local ambiguity in measuring the motion of contour segments (called the "aperture problem") forces us to rely on measurements along the normal direction. Since the measured "normal flow" itself does not agree with the actual normal flow, the "full flow" recovered from this erroneous normal flow possesses substantial error too, and any attempt to recover the 3-D motion from such full flow is doomed to failure. Our method is based on the observation that a polynomial approximation of image flow provides sufficient information for 3-D motion computation. The use of an explicit flow model enables us to improve normal flow estimates through an iterative process. We discuss the adequacy and the convergence of the proposed algorithm. The algorithm has been tested on synthetic and some of simple natural time-varying images. The image flow recovered from this scheme was sufficiently accurate so as to be useful in 3-D structure and motion computation.