ISPRS Commission II, Vol.34, Part 3A „Photogrammetric Computer Vision‘, Graz, 2002
(a) Initial vector (b) After 35 iterations
(c) Final result
Experimental results on synthetic images
The approach was then experimented on Landsat 7 images
using 7 bands (30m resolution, we ignored band 6) to update
features of the Canadian National Topographic Database
(NTDB). The presented result in figure 2 show a homogeneous
water region. The vector initializing the snake is shown in
white on image (a) over enhanced band 2 of the Landsat7
image. Notice the lighter region inside the water area due to
dense vegetation covering part of the lake and mostly obvious
on this specific band. The mixture evaluation clearly defined
two textures for this type of area that initially seems to be
homogeneous on other bands. Over the enhanced image of
band 4, an intermediate result of the snake process is presented
on image (b) and the final result is presented on image (c). The
specific parameters values for this result are the same as the
previous example.
RTE
(a) Initial vector (b) After 70 iterations
Figure 2. Experimental results for water region on
Landsat 7 images
(c) Final result
Finally, results of the experiment on 7 bands of Landsat 7
images for textured regions like vegetation are shown in Figure
3. Notice that is difficult to visually delineate this kind of area
on any of the bands. We present the result on an enhanced
image of band 8 (15m resolution). The initial snake is
presented in image (a), an intermediate image of the snake
process is shown on image (b) and image (c) presents the final
result. The only specific parameter different from the previous
examples is the decreasing factor, À , that was set to 0.7.
(b) After 40 iterations
Figure 3. Experimental results for vegetation area on
Landsat 7 images
(c) Final result
(a) Initial vector
7. CONCLUSION
Our work demonstrates the importance of using snakes and
multi-spectral images for updating existing spatial area
information. We propose a promising automatic approach for
the update of existing vectors in topographic databases. The
combined statistics and gradient information for the external
energy allows the snake to grow in both directions and
provides complementary measures to precisely guide its
deformation. Indeed, the proposed weighted MAP estimation
strategically takes advantages of their respective strengths and
overcomes their drawbacks. An efficient implementation using
finite elements has been proposed for accurate localization.
Finally, experimental results demonstrate the reliability of the
approach.
8. REFERENCES
Auclair-Fortier, M.F., Ziou, D., Armenakis C., and Wang., S.,
2001. Automated Correction and Updating of Road Databases
from High-Resolution Imagery. Canadian Journal of Remote
Sensing, vol. 27, no. 1, pp. 76-89.
Bentabet, L., S. Jodouin, D. Ziou and J. Vaillancourt, 2001.
Automated Updating of Road Databases from SAR Imagery:
Integration of Road Databases and SAR Imagery information.
Proceedings of the Fourth International Conference on
Information Fusion, vol. WeAl, pp. 3-10.
C.A. Bouman and K. Sauer, “A Generalized Gaussian Image
Model for Edge-Preserving MAP Estimation”, IEEE
Transactions on Image Processing, vol. 2, no. 3, pp. 296-310,
1993.
Chakraborty, A., and Duncan, J.S., 1999. Game-Theoretic
Integration for Image Segmentation. IEEE Transactions on
Pattern Analysis and Machine Intelligence, vol. 21, no. 1, pp.
12-30.
Drewniok, C., 1994. Multi-Spectral Edge Detection: Some
Experiments on Data from Landsat-TM. International Journal
of Remote Sensing, vol. 15, no. 18, pp. 3743-3765.
Horritt, M.S., 1999. A statistical active contour model for SAR
image segmentation. /mage and Vision Computing, vol. 17, pp.
213-224.
Kass, M., Witkin, A., and Terzopoulos, D., 1988. Snakes:
Active Contour Models. International Journal of Computer
Vision, vol. 1, no. 4, pp. 321-331.
Olivier, C., Jouzel F., and El Matouat, A., 1999. Choice of the
number of component clusters in mixture models by
information. Vision Interface, pp. 74-78.
Pavlidis, T., and Liow, Y., 1990. Integrating Region Growing
and Edge Detection. IEEE Transactions on Pattern Analysis
and Machine Intelligence, vol. 12, pp. 225-233.
Saporta, G., 1990. Probabilité analyse de données et
statistique. Édition Technip, ISBN: 2-7108-0565-0.
Tek, H. and Kimia, B.B., 1995. Image Segmentation by
Reaction Diffusion Bubbles. Proceedings of the International
Conference On Computer Vision (ICCV), pp. 156-162.
Voorhees, H., and Poggio, T., 1987. Detecting Textons And
Texture Boundaries in Natural Images. International
Conference On Computer Vision (ICCV), pp. 250-258.
A - 186