Abstract
Low-rank matrix approximation has applications in many fields, such as 3D reconstruction from an image sequence and 2D filter design. In this paper, one issue with low-rank matrix approximation is re-investigated: the missing data problem. Much effort was devoted to this problem, and the Wiberg algorithm or the damped Newton algorithm were recommended in previous studies. However, the Wiberg or damped Newton algorithms do not suit for large (especially “long”) matrices, because one needs to solve a large linear system in every iteration. In this paper, we revitalize the usage of the Levenberg-Marquardt algorithm for solving the missing data problem, by utilizing the property that low-rank approximation is a minimization problem on subspaces. In two proposed implementations of the Levenberg-Marquardt algorithm, one only needs to solve a much smaller linear system in every iteration, especially for “long” matrices. Simulations and experiments on real data show the superiority of the proposed algorithms. Though the proposed algorithms achieve a high success rate in estimating the optimal solution by random initialization, as illustrated by real examples; it still remains an open issue how to properly do the initialization in a severe situation (that is, a large amount of data is missing and with high-level noise).
Similar content being viewed by others
References
Aguiar, P. M. Q., & Moura, J. M. F. (2003). Rank 1 weighted factorization for 3D structure recovery: algorithms and performance analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 25(9), 1134–1149.
Aguiar, P. M. Q., & Moura, J. M. F. (1999). Factorization as a rank 1 problem. In Proc. conf. computer vision and pattern recognition (pp. 178–184).
Aguiar, P. M. Q., & Moura, J. M. F. (2000). Weighted factorization. In ICIP (pp. 549–552).
Anandan, P., & Irani, M. (2002). Factorization with uncertainty. International Journal of Computer Vision, 49(2/3), 101–116.
Brand, M. (2002). Incremental singular value decomposition of uncertain data with missing values. In Proc. European conf. computer vision (pp. 707–720).
Brandt, S. (2002). Closed-form solutions for affine reconstruction under missing data. In Statistical methods for video processing workshop, in conjunction with ECCV02 (pp. 109–114).
Buchanan, A. M., & Fitzgibbon, A. W. (2005). Damped Newton algorithms for matrix factorization with missing data. In Proceedings of IEEE computer vision and pattern recognition.
Chen, P. (2004). An investigation of statistical aspects of linear subspace analysis for computer vision applications. Ph.D. thesis, Monash University.
Chen, P. (2008). Heteroscedastic low-rank matrix approximation by the Wiberg algorithm. IEEE Transactions on Signal Processing 56(4), 1429–1439.
Chen, P., & Suter, D. (2004). Recovering the missing components in a large noisy low-rank matrix: Application to sfm. IEEE Transactions on Pattern Analysis and Machine Intelligence, 26(8), 1051–1063.
Chen, P., & Suter, D. (2007). A bilinear approach to the parameter estimation of a general heteroscedastic linear system, with application to conic fitting. Journal of Mathematical Imaging and Vision, 28(3), 191–208.
Edelman, A., Arias, T. A., & Smith, S. T. (1998). The geometry of algorithms with orthogonality constraints. SIAM Journal on Matrix Analysis and Applications, 20(2), 303–353.
Golub, G. H., & Loan, C. F. V. (1996). Matrix computations (3rd ed.). Baltimore: Johns Hopkins Press.
Guerreiro, R. F. C., & Aguiar, P. M. Q. (2002). 3D structure from video streams with partially overlapping images. In IEEE international conference on image processing ICIP’02.
Guerreiro, R. F. C., & Aguiar, P. M. Q. (2003). Estimation of rank deficient matrices from partial observations: Two-step iterative algorithms. In Energy minimization methods in computer vision and pattern recognition.
Hartley, R. I., & Zisserman, A. (2003). Multiple view geometry in computer vision (2nd ed.). Cambridge: Cambridge University Press.
Heyden, A., & Kahl, F. (1998). Reconstruction from affine cameras using closure constraints. In ICPR (pp. 47–50).
Hua, Y., Nikpour, M., & Stoica, P. (2001). Optimal reduced-rank estimation and filtering. IEEE Transactions on Signal Processing, 49(3), 457–469.
Irani, M., & Anandan, P. (2000). Factorization with uncertainty. In Proc. European conf. computer vision (Vol. 49, pp. 539–553).
Jacobs, D. (1997). Linear fitting with missing data: Applications to structure-from-motion and to characterizing intensity images. In Proc. conf. computer vision and pattern recognition (pp. 206–212).
Jacobs, D. (2001). Linear fitting with missing data for structure-from-motion. Computer Vision and Image Understanding, 82, 57–81.
Kahl, F., & Heyden, A. (1998). Structure and motion from points, lines and conics with affine cameras. In Proc. European conf. computer vision (pp. 327–341).
Kahl, F., & Heyden, A. (1999). Affine structure and motion from points, lines and conics. International Journal of Computer Vision, 33(3), 163–180.
De la Torre, F., & Black, M. J. (2003). A framework for robust subspace learning. International Journal of Computer Vision, 54(1–3), 117–142.
Lu, W. S., Pei, S. C., & Wang, P. H. (1997). Weighted low-rank approximation of general complex matrices and its application in the design of 2D digital filter. IEEE Transactions on Circuits and Systems I, 44, 650–655.
Manton, J. H., Mahony, R., & Hua, Y. (2003). The geometry of weighted low-rank approximations. IEEE Transactions on Signal Processing, 51(2), 500–514.
Martinec, D., & Pajdla, T. (2002). Structure from many perspective images with occlusion. In Proc. European conf. computer vision (pp. 355–369).
Martinec, D., & Pajdla, T. (2005). 3D reconstruction by fitting low-rank matrices with missing data. In Proc. conf. computer vision and pattern recognition (pp. 198–205).
Morris, D. D., & Kanade, T. (1998). A unified factorization algorithm for points, line segments and planes with uncertainty models. In Proc. int. conf. computer vision (pp. 696–702).
Nocedal, J., & Wright, S. J. (1999). Numerical optimization. Berlin: Springer.
Okatani, T., & Deguchi, K. (2006). On the Wiberg algorithm for matrix factorization in the presence of missing components. International Journal of Computer Vision, 72(3), 329–337.
Poelman, C., & Kanade, T. (1997). A paraperspective factorization method for shape and motion recovery. IEEE Transactions on Pattern Analysis and Machine Intelligence, 19(3), 206–219.
Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1992). Numerical recipes in C (2nd ed.). Cambridge: Cambridge University Press.
Shum, H., Ikeuchi, K., & Reddy, R. (1995). Principal component analysis with missing data and its applications to polyhedral object modeling. IEEE Transactions on Pattern Analysis and Machine Intelligence, 17(9), 854–867.
Tomasi, C., & Kanade, T. (1992). Shape and motion from image streams under orthography: A factorization method. International Journal of Computer Vision, 9(2), 137–154.
Triggs, B., McLauchlan, P., Hartley, R., & Fitzgibbon, A. (2000). Bundle adjustment—A modern synthesis. In W. Triggs, A. Zisserman, & R. Szeliski (Eds.), LNCS. Vision algorithms: theory and practice (pp. 298–375). Berlin: Springer.
Vidal, R., & Hartley, R. (2004). Motion segmentation with missing data using power factorization and gpca. In Proc. conf. computer vision and pattern recognition.
Wiberg, T. (1976). Computation of principle components when data are missing. In 2nd symposium computational statistics.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Chen, P. Optimization Algorithms on Subspaces: Revisiting Missing Data Problem in Low-Rank Matrix. Int J Comput Vis 80, 125–142 (2008). https://doi.org/10.1007/s11263-008-0135-7
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11263-008-0135-7