determined to be 1;;75pixel, h;;715pixel, and Aq? = 0.79. From 
this information we can conclude that the two stamps are nearly 
perfectly aligned regarding the orientation but only poorly 
aligned regarding the position particularly in the direction per- 
pendicular to the direction of the script. It took about 20 ms on a 
2 GHz Pentium 4 to search the print in the entire image of size 
652x494 allowing a 360? rotation. The 20 ms can be completely 
attributed to the search of the root part. The search of the second 
part was too fast to be measured. If we would search the second 
part separately it would also take 20 ms to find it. Therefore, by 
using our hierarchical model the recognition time was reduced 
to 50%. 
In the second example the variations of another writing are 
analyzed (see Figure 10). The resulting search tree in the lower 
right image of Figure 10 visualizes the resulting optimum 
search strategy. The result is to search each letter relatively to 
its neighboring letter, which corresponds to our intuition. 
  
Figure 10. The variations in the writing are analyzed and used to 
build the hierarchical model. The calculated search tree is 
visualized in the lower right image, where each letter is 
searched relative to its neighboring letter, as we would expect. 
It took 60 ms to search the entire hierarchical model in a 
512x512 image allowing a 180° rotation. 50 ms can be 
attributed to the search of the root part and only 10 ms to the 
search of all other parts. In a comparison to a complete search 
of all parts in the entire image (450 ms) this is a reduction to 
1396. 
5. CONCLUSIONS 
In this paper we presented an approach for hierarchical auto- 
matic object decomposition for object recognition. This is use- 
ful when searching for objects that consist of several parts that 
can move relative to each other, which often happens in 
industry, for example. A hierarchical model was automatically 
created using several example images that show the relative 
movements of the single parts. This model can be utilized to 
efficiently search for the object in an arbitrary image. The 
examples shown in section 4 emphasize the high potential of 
our novel approach. 
REFERENCES 
Bajcsy, R. Kovacic, S. 1989. Multi-resolution elastic 
matching. Computer Vision, Graphics, and Image Processing, 
46(1): pp. 1-21. 
Ballard, D. H., 1981. Generalizing the Hough transform to 
detect arbitrary shapes. Pattern Recognition, 13(2), pp. 111-122. 
Borgefors, G., 1988. Hierarchical chamfer matching: A 
parametric edge matching algorithm. IEEE Transactions on 
Pattern Analysis and Machine Intelligence, 10(6), pp. 849-865. 
Brown, L. G., 1992. A survey of image registration techniques. 
ACM Computing Surveys, 24(4), pp. 325-376. 
Chu, Y. J. and Tseng-Hong, L., 1965. On the shortest 
arborescence of a directed graph. Scientia Sinica, 14(10), pp. 
1396-1400. 
Elder, J., 1999. *Are Edges Incomplete?". International Journal 
of Computer Vision, 34(2/3), pp. 97-122. 
Hauck, A. Lanser, S. and Zierl, C., 1997. Hierarchical 
Recognition of Articulated Objects from Single Perspective 
Views. In: Proc. Computer Vision and Pattern Recognition 
(CVPR '97), IEEE Computer Society Press, pp. 870-883. 
Huttenlocher, D. P., Klanderman, and G. A., Rucklidge, W. J., 
1993. Comparing Images Using the Hausdorff Distance. IEEE 
Transactions on Pattern Analysis and Machine Intelligence, 
15(9), pp. 850-863. 
Jain, A. K., Zhong, Y., and Lakshmanan, S., 1996. Object 
matching using deformable templates. JEEE Transactions on 
patterns analysis and machine intelligence, 18(3), pp. 267-277. 
Koch, K. R., 1987. Parameterschátzung und Hypothesentests in 
linearen Modellen. Dümmler, Bonn. 
Koffka, K., 1935. Principles of Gestalt Psychology. Harcourt 
Brace, New York. 
Lai, S. and Fang, M., 1999. Accurate and fast pattern 
localization algorithm for automated visual inspection. Real- 
Time Imaging, 5, pp. 3-14. 
Lowe, D.G., 1985. Perceptual Organization and Visual 
Recognition, Kluwer Academics, Boston. 
Marr, D., 1982. Vision, W.H. Freeman and Company, San 
Francisco, CA. 
Rucklidge, W. J., 1997. Efficiently locating objects using the 
Hausdorff distance. International Journal of Computer Vision, 
24(3), pp. 251-270. 
Steger, C., 2001. Similarity measures for occlusion, clutter, and 
illumination invariant object recognition. In: Mustererkennung 
2001, B. Radig and S. Florczyk (eds), Springer, Berlin, pp. 148- 
154. 
Ullman, S., 1979. The interpretation of visual motion. MIT 
Press, Cambridge, MA. 
Ulrich, M., Steger, C., Baumgartner, A., and Ebner, H., 2001. 
Real-time object recognition in digital images for industrial 
applications. In: 5" Conference on Optical 3-D Measurement 
Techniques, Vienna, Austria, pp. 308-318. 
Ulrich, M. and Steger, C., 2001. Empirical performance 
evaluation of object recognition methods. In: Empirical 
Evaluation Methods in Computer Vision, H. I. Christensen and 
P. J. Phillips (eds), IEEE Computer Society Press, Los 
Alamitos, CA, pp. 62-76. 
Wertheimer, M., 1938. Laws of Organization in Perceptual 
Forms. In: A Source Book of Gestalt Psychology, W. D. Ellis 
(ed), Harcourt Brace. 
Witkin, A.P. and Tenenbaum, J.M., 1983. On the Role of 
Structure in Vision. In: Human and Machine Vision, Jacob Beck 
and Barbara Hope and Azriel Rosenfeld (eds), Academic Press, 
New York. 
-]104— 
KEY 
ABS] 
In the 
incree 
been 
adequ 
Thus, 
projec 
recon 
photo 
artific 
made 
are di 
KUR 
Im v 
welcl 
Anwt 
geme 
Allge 
Daru 
Ober 
Ansa 
Beid 
glatte 
das z 
inter] 
In tl 
appli 
a ma 
used 
obje: 
achie 
The 
stron 
task 
requ 
reso) 
othe 
  
Co