3D reconstruction method with 1D subspace

To recover 3D objects from a sequence of images, a 3D reconstruction method is presented based on 1D subspace under the orthographic projection assumption. The method relies on the facts that the row vectors composed of all the image points span, the same linear subspace as the row vectors composed of 3D space points span and that a basis of the subspace can consist of two row vectors composed of all the image points and one row vector in 3D space that is orthogonal to the former. The row vector can be determined, and the 3D reconstruction is accomplished. The novelty lies in the fact that the method can treat all the image points uniformly. Experiments with both simulation data and real data show that the method is efficient and robust with small re-projection errors.