IMAGE DESCRIPTION AND 3-D RECONSTRUCTION FROM IMAGE TRAJECTORIES OF ROTATIONAL MOTION

This paper presents a new technique for reconstructing the 3-D structu re and motion of a scene undergoing relative rotational motion with re spect to the camera. Given image correspondences of point features tra cked over many frames, a two-stage technique for reconstruction is pre sented. First, a grouping algorithm that exploits spatio-temporal cons traints of the common motion to achieve a reliable description of disc rete point correspondences as curved trajectories (general conics in t he case of rotational motion) in the image plane is developed. In cont rast, trajectories fitted to points independent of each other lead to arbitrary image descriptions and very inaccurate 3-D parameters. Secon d, a new closed-form solution, under perspective projection, for the 3 -D motion and location of points from the computed image trajectories is presented. Both stages are applied to real image sequences with goo d results. This approach represents a first step in a longer-term rese arch effort examining the role of explicit spatio-temporal organizatio n in the interpretation of scenes from dynamic images.