Structure from Motion

Structure from motion (SfM) refers to the process of automatically generate a 3D structure of an object by its tracked 2D image frames. Suppose we are given trajectories of P tracked feature points of a rigid object from F 2D frames of a rigidly moving camera. Then, finding the full 3D reconstruction of this object can be posed as a low-rank matrix recovery problem.

Publication

R. Liu, Z. Lin, Z. Su, "Exactly Recovering Low-Rank Matrix in Linear Time via l1 Filter", August 2011.

L. Wu, Y. Wang, Y. Liu, Y. Wang, “Robust structure from motion with affine camera via low-rank matrix recovery”, China Information Sciences, Volume 56, Issue 11, pages 1-10, November 2013.

F. Arrigoni, L. Magri, B. Rossi, P. Fragneto, A. Fusiello, “Robust Absolute Rotation Estimation via Low-rank and Sparse Matrix Decomposition”, International Conference on 3D Vision, 3DV 2014, pages 491-498, 2014.

F. Arrigoni, B. Rossi, A. Fusiello, "Robust and Efficient Camera Motion Synchronization via Matrix Decomposition", International Conference on Image Analysis and Processing, ICIAP 2015, September 2015.

F. Arrigoni, B. Rossi, P. Fragneto, A. Fusiello, “Robust Synchronization in SO(3) and SE(3) via Low-rank and Sparse Matrix Decomposition”, Computer Vision and Image Understanding, Volume 174, pages 95-113, 2018.

R. Kennedy, L. Balzano, S. Wright, C. Taylor, “Online algorithms for factorization-based structure from motion”, WACV 2014, 2014.

R. Kennedy, L. Balzano, S. Wright, C. Taylor, “Online algorithms for factorization-based structure from motion”, Computer Vision and Image Understanding, Volume 150, pages 139-152, 2016.

X. Wang, F. Wang, Y. Chen, "Capturing Complex 3D Human Motions with Kernelized Low-Rank Representation from Monocular RGB Camera", MDPI Sensors 2017, Volume 17, No. 9, 2017.

C. Olsson, M. Carlsson, E. Bylow, “A Non-Convex Relaxation for Fixed-Rank Approximation”, IEEE International Workshop on RSL-CV in conjunction with ICCV 2017, October 2017.