I. Introduction
The term Global Motion (GM) is used to describe the coherent component of motions of different constituent parts of an object, by a parameterized motion model. The process of estimating these parameters is known as Global Motion Estimation (GME). Commonly, global motion refers to the apparent background motion, induced by a moving camera. Global motion estimation is used in several applications viz. image stabilization [2], [3] which is discussed in details in this paper, video compression (sprite coding and global motion compensation in MPEG-4) [4], foreground-background segmentation, mosaicing and image registration [5]. The parametric models used include 2-parameter translation model, 4-parameter rotation-scale-translation (RST) model, 6-parameter affine model, 8-parameter projective model, 8-parameter bilinear model etc. [6]. Translation and RST model parameters are easier to estimate but they don't have enough flexibility to adequately describe the global motion pattern in image sequences. On the other hand, projective and bilinear models offer a great deal of flexibility in terms of their ability to describe complex motion pattern, but the methods for estimating parameters for such models are very complex for real time applications [7]. Moreover, increased representational capability of the model doesn't always result in better representation, since the model starts fitting noise. Because of these considerations, among these methods affine motion model is very popular as it provides a tradeoff between generality and ease of estimation [8], [9].