Image deformations are better than optical flow

In many computer vision applications it is necessary to compute the direction of heading of a moving camera from the images it produces. Traditionally, this computation has been based on the optical flow, that is, on the motion of point features in the field of view. We show that the differential changes in the angles between the projection rays of pairs of point features are a better input for this purpose. These angles, the image deformations, do not depend on viewer rotation, so the key problem of separating the effects of rotation from those of translation is solved at the input. Experiments show both the feasibility of the method on real images and the advantages of using deformations rather than optical flow.

Duke Scholars

Author Carlo Tomasi Computer Science