An Iterative Multiresolution Scheme for SFM
Several factorization techniques have been proposed for tackling the Structure from Motion problem. Most of them provide a good solution, while the amount of missing and noisy data is within an acceptable ratio. Focussing on this problem, we propose to use an incremenal multiresolution scheme, with classical factorization techniques. Information recovered following a coarse-to-fine strategy is used for both, filling in the missing entries of the input matrix and denoising original data. An evaluation study, by using two different factorization techniques–the Alternation and the Damped Newton–is presented for both synthetic data and real video sequences.
KeywordsFeature Point Singular Value Decomposition Factorization Technique Input Matrix Structure From Motion
Unable to display preview. Download preview PDF.
- 2.Tomasi, C., Kanade, T.: Shape and motion from image streams: a factorization method. Full report on the orthographic case (1992)Google Scholar
- 3.Jacobs, D.: Linear fitting with missing data for structure-from-motion. In: Computer vision and image understanding, CVIU, pp. 7–81 (2001)Google Scholar
- 5.Brandt, S.: Closed-form solutions for affine reconstruction under missing data. In: Proceedings Statistical Methods for Video Processing Workshop, in conjunction with ECCV, pp. 109–114 (2002)Google Scholar
- 6.Chen, P., Suter, D.: Recovering the missing components in a large noisy low-rank matrix: Application to sfm. IEEE Transactions on Pattern Analysis and Machine Intelligence 26 (2004)Google Scholar
- 7.Buchanan, A., Fitzgibbon, A.: Damped newton algorithms for matrix factorization with missing data. In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), vol. 2, pp. 316–322 (2005)Google Scholar
- 9.Ma, Y., Soatto, J., Kosecká, J., Sastry, S.: An invitation to 3d vision: From images to geometric models. Springer, New York (2004)Google Scholar