Motion analysis is central to many applications where time-varying image data are involved, such as robotics and multimedia systems. This thesis describes a new motion analysis framework using multiresolution affine models. Traditional approaches using local translational models lack sufficient flexibility to derive robust and concise descriptions of motion, since they are only valid over small local regions. Affine models, however, provide a closer approximation to regional motion variation, hence allowing for larger analysis windows to be employed, leading to greater robustness and conciseness. The new framework is applied to the motion estimation and tracking problems. The estimator employs an affine model of local motion and the model parameters are estimated using generalised correlation, implemented efficiently in the frequency domain. A multiresolution framework allows the approach to adapt to the data by varying the analysis window size. Regions where the motion has high spatial variance can then be represented accurately, whereas regions of near constant motion can be described with low overhead. Combining a multiresolution technique and an affine motion model leads to a robust and concise motion description. The estimator is shown to perform well over a range of different image sequences exhibiting different types of motions. The estimation framework is then generalised into the temporal domain. A novel region representation, closed under affine motion, is proposed, using 2-D Gaussians organised into a binary tree. This \emphg-blob representation describes a sequence as a set of connected elements corresponding to projections of planar surface patches in 3-D motion. This description is concise and is therefore an ideal framework for tracking the temporal evolution of regions. A tracker is implemented using a Kalman filter attached to the velocity component of each g-blob where the Kalman filter is modified to take into account rapidly changing motion trajectories. The region tracker performs well, tracking successfully the evolution of regions undergoing intricate motions.