ation
jority
point
s in-
TEES.
„ the
hows
r real
when
'sults
pairs
hows
those
it im-
| col-
mn 3
used
ind 4
yr re-
| im-
It is
good
iding
5 and
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Object total matches | matches used for | inliers | outliers
epipolar geometry
aerial image pair | 67 67 39 8
aerial image pair 2 51 51 42 9
turntable images 'Obi' 229 68 66 2
virtual turntable images 282 84 80 4
Valbonne image pair 112 50 41 9
Table 1: Evaluation of the matching performance. Results are given for 5 image pairs. Note that for the turntable images
as well as for the virtual turn table scene most of the inliers lie inside a planar region, for the aerial image pairs several
matches lie on depth-discontinuities where the Zwickel-based descriptor is well suited. For the Valbonne image pair
several matches were found at depth discontinuities since many prominent lines were found on the borders of planar
regions.
14 seconds on a Pentium 4 machine with 2.4 GHz.
5 CONCLUSION
We described a novel approach for computing affine invari-
ant descriptors from Zwickels. Our experiments show, that
these descriptors are invariant against viewpoint changes
as well as illumination changes. Our method is suitable
for images where a sufficient number of lines and therefore
Zwickels can be extracted and the sectors inside the Zwick-
els provide enough texture information to distinguish com-
peting candidates. Further possible improvements are the
use of more complex distance measures for histogram com-
parison, such as the earth movers distance. In the next set
of experiments we plan to test the method also in an object
recognition context.
ACKNOWLEDGMENTS
This work has been done in the VRVis research center,
Graz and Vienna/Austria (http://www.vrvis.at), which is
partly funded by the Austrian government research pro-
gram Kplus. Horst Bischof acknowledges the support of
the Kplus competence center Advanced Computer Vision
(ACV) funded by the Kplus program. The authors wish
to thank Sandra Ober for providing the turntable data and
Konrad Schindler for providing the virtual turntable se-
quence.
REFERENCES
Baumberg, A., 2000. Reliable feature matching across
widely separated views.
Canny, J., 1986. A computational approach to edge detec-
tion. IEEE Transactions on Pattern Analysis and Machine
Intelligence, 8(6):679-698. 8(6), pp. 679-698.
Harris, C. and Stephens, M., 1988. A combined corner and
edge detector. Proceedings 4th Alvey Visual Conference.
Huet, B. and Hancock, E., 1996. Cartographic indexing
into a database of remotely sensed images. In: WACV’96,
pages 8-14, Dec 1996.
Lowe, D. G., 1999. Object recognition from local scale-
invariant features. In: Proc. of the International Confer-
ence on Computer Vision ICCV, Corfu, pp. 1150-1157.
Matas, J., Chum, O., Urban, M. and Pajdla, T., 2002. Ro-
bust wide baseline stereo from maximally stable extremal
regions. In: Proc. 13th British Machine Vision Conference,
Cardiff, UK, pp. 384—393.
Mikolajczyk, K. and Schmid, C., 2001. Indexing based
on scale invariant interest points. In: Proceedings of the
International Conference on Computer Vision, Vancouver,
Canada, pp. 525-531.
Mikolajczyk, K. and Schmid, C., June 2003. A perfor-
mance evaluation of local descriptors. In: International
Conference on Computer Vision and Pattern Recognition
(CVPR'2003), Vol. 2, pp. 257-263.
Rubner, Y., Tomasi, C. and Guibas, L. J., 1998. A metric
for distributions with applications to image databases. In:
Proceedings of the 1998 IEEE International Conference on
Computer Vision, Bombay, India, January 1998, pp. 59-
66.
Schaffalitzky, F. and Zisserman, A., 2001. Viewpoint in-
variant texture matching and wide baseline stereo. In:
Proc. 8th International Conference on Computer Vision,
Vancouver, Canada.
Schmid, C. and Mohr, R., 1997. Local grayvalue invariants
for image retrieval. IEEE Transactions on Pattern Analysis
and Machine Intelligence 19(5), pp. 530—535.
Tuytelaars, T. and Gool, L. V., 2000. Wide baseline stereo
matching based on local, affinely invariant regions. In:
British Machine Vision Conference BMVC'2000.
