Building Computer Vision Projects with OpenCV 4 and C++

Affine constraint

For a robust and meaningful pairwise matching, we often apply a geometric constraint. One such constraint can be an affine transform, a transform that allows only for scale, rotation, and translation. In 2D, an affine transform can be modeled in a 2 x 3 matrix:

To impose the constraint, we look for an affine transform  that minimizes the distance (error) between matching points from the left  and right  images.