3D Structure Recovery

Recovering 3D structure from motion in noisy 2D images is a fundamental problem addressed by computer vision researchers. By consistently tracking feature points across multiple images using the Lucas-Kanade optical flow algorithm, the Tomasi-Kanade factorization method enables 3D shape reconstruction through matrix decomposition of the feature point trajectories.