The matching problem is to find corresponding points between images when the imaging conditions
significantly change. We focus on the case when the resolution of images is very different.
Given corresponding couples of points we can determine the transformation between two images.
If the images represent planar scenes we can estimate a 3x3 homography matrix. In the case of 3D scenes
a 3x3 fundamental matrix can be computed using epipolar geometry principles.
Move the mouse pointer over the images below.

To compute the transformation matrix we need to establish several point-to-point correspondences.
Stereo Vision applet illustrates the problem and estimates the transformation matrix.