“…GME techniques differ as far as the parametric model of camera motion is concerned and in the way the model parameters are estimated. The following motion models have been used in the past: 2-parameter model corresponding to pan [2], [3], 3-parameter model corresponding to pan and zoom [4], 4-parameter linear model corresponding to pan and zoom [5], 4-parameter similarity model corresponding to pan, zoom and rotation, 6-parameter affine model [6,7,[9][10][11], 8-parameter perspective model [12][13][14][15][16]. There are two main categories of method for estimating global motion: block-based twostage GME algorithms [4][5][6], and pixel-based gradient descent (GD) GME algorithms [1,7,[10][11][12][13]16,17].…”