?D features
e proposed
reconstruc-
issing data
D features
The swap-
atched fea-
erved.
implement
lds consid-
her degree
to the pro-
ed method
t was con-
ne scheme
rging runs
sult clearly
:thod.
ISPRS Commission III, Vol.34, Part 3A ,,Photogrammetric Computer Vision", Graz, 2002
>
Cc
| | M Christy-Horaud
[] Proposed Method
| a :
Qt 2 3-45" 6 7T 8 9 1)
Percent Errors
Mean Error (Pixels)
ed — Do Na to GO >
ce en ce e e Ce oo
Um
Figure 12: Percentage of 2D features swapped to emu-
lated errors in the correspondence algorithm, and the cor-
responding mean error of the reconstruction.
m Original Method
|__| @Proposed Method
Number of Non-Convergence
0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-40
Bins of Percent Error
Figure 13: The number of non-converging runs with in-
creasing number of errors, with and without the proposed
method for Eucledian reconstuction. The runs are pooled
in bins of 5.
6 DISCUSSION
Factorization algorithms represent an important set of tools
for recovering structure and motion from an image stream.
They can be used as a full solution to the problem or as the
very important initialization step in non-linear minimiza-
tion methods [Slama, 1984, Triggs et al., 2000].
We have presented a computationally efficient algorithm
for applying arbitrary error functions in the factorization
scheme for structure and motion. It has also been demon-
started on real typical data and via rigorous tests, that this
scheme deals well with erroneous data.
It is noted, that the particular choice of error function is up
to the fancy of the user. For a further survey of the benefits
of different error functions the reader is referred to [Black
and Rangarajan, 1996].
REFERENCES
Black, M. and Rangarajan, A., 1996. On the unification of
line processes, outlier rejection, and robust statistics with
applications in early vision. International Journal of Com-
puter Vision 19(1), pp. 57-91.
Booker, A., Dennis, J., Frank, P., Serafini, D., Torczon,
V. and Trosset, M., 1999. A rigorous framework for opti-
mization of expensive functions by surrogates. Structural
Optimization 17(1), pp. 1-13.
Christy, S. and Horaud, R., 1994. Euclidian shape and mo-
tion from multiple perspective views by affine iterations.
Technical Report 2421, INRIA.
Christy, S. and Horaud, R., 1996. Euclidean shape and
motion from multiple perspective views by affine itera-
tion. IEEE Trans. Pattern Analysis and Machine Intelli-
gence 18(11), pp. 1098-1104.
Costeira, J. and Kanade, T., 1998. A multibody factor-
ization method for independently moving objects. Int'l J.
Computer Vision’98 29(3), pp. 159-179.
Fletcher, R., 1987. Practical Methods of Optimizations.
John Wiley & Sons.
Harris, C. and Stephens, M., 1988. A combined corner and
edge detector. In: Proc. Alvey Conf., pp. 189-192.
Hartley, R. and Zisserman, A., 2000. Multiple View Geom-
etry. Cambridge University Press, The Edinburgh Build-
ing, Cambridge CB2 2RU, UK.
Huber, P., 1981. Robust Statistics. John Wiley, New York.
Irani, M. and Anandan, P., 2000. Factorization with un-
certainty. In: European Conf, Computer Vision'2000,
pp. 539—553.
Jacobs, D., 2001. Linear fitting with missing data for
structure-from-motion. Computer Vision and Image Un-
derstanding 82(1), pp. 57-81.
