We focus in this paper on the problem of adding computer-generated objects in video sequences that have been shot with a zoom lens camera. While numerous papers have been devoted to registration with fixed focal length, little attention has been brought to zoom lens cameras. In this paper, we propose an efficient two-stage algorithm for handling zoom changing which are are likely to happen in a video sequence. We first attempt to partition the video into camera motions and zoom variations. Then, classical registration methods are used on the image frames labeled camera motion while keeping the internal parameters constant, whereas the zoom parameters are only updated for the frames labeled zoom variations. Results are presented demonstrating registration on various sequences. Augmented video sequences are also shown.