In this paper we present a new method to estimate optical flow for large displacements. It is based on prediction of global flow field parameters, performs better than multiresolution estimation methods and has been verified using standard test sequences as well as real-world data. Global flow field parameters can be estimated from optical flow measurements in all flow regions. They can then be used to predict the flow in optical flow regions of the next frame. This technique reduces the complexity in comparison to hierarchy-based methods, while the flow field parameters can also be used to compensate optical flow produced by egomotion. KEY WORDS Motion Detection and Estimation, Multi-Resolution and Multi-Spectral Processing, Optical Flow, Vision