Using priors for improving generalization inNon-Rigid structure-from-motion
Research output: Contribution to conference › Paper › Research › peer-review
Standard
Using priors for improving generalization inNon-Rigid structure-from-motion. / Olsen, S. I.; Bartoli, A.
2007. Paper presented at 2007 18th British Machine Vision Conference, BMVC 2007, Warwick, United Kingdom.Research output: Contribution to conference › Paper › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - CONF
T1 - Using priors for improving generalization inNon-Rigid structure-from-motion
AU - Olsen, S. I.
AU - Bartoli, A.
PY - 2007
Y1 - 2007
N2 - This paper describes how the generalization ability of methods for non-rigid Structure-from-Motion can be improved by using priors. Most point tracks are often visible only in some of the images; predicting the missing data can be important. Previous Maximum-Likelihood (ML)-approaches on implicit non-rigid Structure-from-Motion generalize badly. Although the estimated model fits well to the visible training data, it often predicts the missing data badly. To improve generalization we propose to add a temporal smoothness prior and a continuous surface shape prior to an ML-approach. The temporal smoothness prior constrains the camera trajectory and the configuration weights to behave smoothly. The surface shape prior constrains consistently close image point tracks to have a similar implicit structure. We propose an algorithm for achieving a Maximum A Posteriori (MAP)-solution and show experimentally that the MAP-solution generalizes far better than the MLsolution. The proposed method is fully automatic: it handles a substantial amount of missing data as well as outlier contaminated data, and automatically estimates the rank of the measurement matrix.
AB - This paper describes how the generalization ability of methods for non-rigid Structure-from-Motion can be improved by using priors. Most point tracks are often visible only in some of the images; predicting the missing data can be important. Previous Maximum-Likelihood (ML)-approaches on implicit non-rigid Structure-from-Motion generalize badly. Although the estimated model fits well to the visible training data, it often predicts the missing data badly. To improve generalization we propose to add a temporal smoothness prior and a continuous surface shape prior to an ML-approach. The temporal smoothness prior constrains the camera trajectory and the configuration weights to behave smoothly. The surface shape prior constrains consistently close image point tracks to have a similar implicit structure. We propose an algorithm for achieving a Maximum A Posteriori (MAP)-solution and show experimentally that the MAP-solution generalizes far better than the MLsolution. The proposed method is fully automatic: it handles a substantial amount of missing data as well as outlier contaminated data, and automatically estimates the rank of the measurement matrix.
UR - http://www.scopus.com/inward/record.url?scp=84898426785&partnerID=8YFLogxK
U2 - 10.5244/C.21.105
DO - 10.5244/C.21.105
M3 - Paper
AN - SCOPUS:84898426785
T2 - 2007 18th British Machine Vision Conference, BMVC 2007
Y2 - 10 September 2007 through 13 September 2007
ER -
ID: 392568462