Can any one please define or share link of affine model with six parameter to describe the image global motion??

Global motion compensation is an essential part of moving target detection in dynamic background.adopt affine model with six parameters to describe the image global motion.this model widely used because of its invariance under motion and variation. such as translation, scaling and 2-D rotation.

Answer 1

[Marco Marcon]

An Affine transformation preserves parallel lines but not angles, and it is usually adopted as an approximation for projective transform when parallel lines almost preserve their parallelism. The model is based on 6 parameters and, since points at infinity stay at infinity while finite points remain finite points, we can assume that its transform is:

Where x, the 3 elements vector on the right, is expressed in 2D uniform coordinates with the last value equal to 1. While the output x' is in inhomogeneous coordinates. 3 corresponding 2D points between two images are sufficient to define the 6 parameters of the transformation and it can be considered a generalization of the scaling transform where the parameters are only 4 (1 for scaling, 2 for translation and one for rotation). The scaling transform:

An exemple for the affine transform is the following:

[Rutvi Modi]

Can you please post code to get homography matrix(in order to obtain the accurate parameters of affine transformation model).