An image processing apparatus includes a first processing part selecting motion detection points for detecting a motion between images, and acquiring first motion vectors that are local motion vectors at the motion detection points, a second processing part acquiring first representative motions that are representative motions of translation, scaling and rotation based on the first motion vectors, a third processing part eliminating components of the first representative motions from the first motion vectors and acquiring second motion vectors, a fourth processing part acquiring local foreshortening motions at the motion detection points from the second motion vectors, and a fifth processing part acquiring a second representative motion that is a representative foreshortening motion from the local foreshortening motions. This process can obtain information representing geometric deformation between the images by simple processing.