Projective Structure and Motion from Image Sequences using Subspace Methods

In this paper a novel method to make reconstruction from any number of uncalibrated cameras is presented. The novelty lies in the fact that the algorithm relies only on subspace methods, which implies that it is independent of the chosen coordinate system in the images. Furthermore, it does not distinguish between the dioeerent points in each image and can deal with any number of images and points in a unified manner. Finally, it is possible to obtain a result that is independent on the chosen ordering of the images. The performance of this iterative algorithm is shown on both simulated and real data. In general, for n points in m images, it is suOEcient to make only 10 iterations, where each iteration involves the calculation of the four eigenvectors corresponding to the largest eigenvalues to an nThetan matrix and calculations of the eigenvector corresponding to the smallest eigenvalue to m dioeerent n Theta n matrices.