— US OS OS
T6 6 6
> Ge
C
ue
he
sts
ird
the
a
'en
ery
al.,
ied
ind
ly
low
S.
ime
sale
31)
orth
dge
EC,
the
| be
urs
hile
with
ped
data
cry,
and
hree
has
International Archives of the Photog,.smmetry, Remote Sensin g and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
^.
E a
c
Fig. 6 Edge mapper. (a) QuickBird, (b) Ikonos, (c) aerial imagery.
The overall characteristics of the colour morphological
operations are similar to that in grayscale case, i.e., the colour
dilation eliminates ‘dark’ details, where the vectors have
smaller rank than its surroundings, enhance ‘light’ details,
where the vectors have larger rank than its surroundings,
reduces ‘dark’ objects and enlarges ‘light’ objects; the colour
erosion eliminates ‘light’ details, enhances ‘dark’ details,
reduces ‘light’ objects, and enlarges ‘dark’ objects; the colour
closing typically eliminates ‘dark’ details, and the colour
opening eliminates ‘light’ objects. The results of the proposed
colour edge detector using simply Euclidean distance as the
vector norm are shown in Fig. 4. All building roofs in the three
scenes have been extracted.
7. DISCUSSION AND OUTLOOK
A novel approach to colour mathematical morphology based on
principal component analysis has been presented. The general
design of colour mathematical morphology is conceptually
sound and the algorithms were tested with phansharpened Im
Ikonos and 61 cm QuickBird satellite images and colour aerial
imagery acquired over a built-up area in Toronto, Ontario. The
proposed method extended the greyscale morphology to the
colour morphology and provided its promising performance in
colour image processing and feature extraction. Preliminary
investigations suggest that building extraction can be
automatically performed by using the developed colour edge
detector. However, we didn’t report those results because of
space limitations. Detailed discussion is being presented in
another publication. In future, the main focus will be on the
roof extraction and aim at a global robust adjustment including
the regularities of roof structure. Automatic procedures may fail
in recovering the correct information due to the complexity of
the task. Therefore, interactive tools for editing the extracted
results are necessary. Future work will also include handling of
road networks using the proposed method within an ongoing
project on automated manmade object extraction from high-
resolution colour satellite imagery. A strategy that integrates
multiple cues including colour and attributed edges in a GIS
environment will be invested and tested.
1173
Acknowledgements:
This research was supported by the Canada Foundation for
Innovation (CFI) and the Natural Science and
Engineering Research Council of Canada (NSERC).
Tony Sani of Spatial Geo-Link, Inc., and Yubin Xin of
PCI Geomatics, Inc., are greatly acknowledged to
provide test images.
References:
Astola. J., Haavisto, P., and Neuvo, Y., 1990. Vector median
filters. Proceedings of IEEE, Vol. 78, pp. 678 — 689.
Barnett, V., 1976. The ordering of multivariate data. Journal of
Royal Statistical Society A, Vol.3, pp. 318 - 355.
Comer, M. L., Del, E. J., 1998. Morphological Operations. The
Colour Image Processing Handbook, Chapman and Hall.
Daniels, H., 1946. The relation between measures of
correlation in the universe of sample permutations,
Biometrika XXXIII, pp. 129-135.
Dougherty, E. R., 1991. Digital Image Processing Methods.
Marcel Dekker, New York.
Goutsias, J., Heijmans, H. J. A. M., and Sivakumar, K., 1995.
Morphological operators for image sequences. ‘omputer
Vision and Image Understanding, Vol. 62, pp. 326 — 346.
Kendall, M., 1962. Rank Correlation Methods, 3/e, Hafner,
New York.
Korhonen, P.,Siljamaki, A., 1998. Ordinal principal component
analysis theory and an application. Computational Statistics
and Analysis, Vol. 26, pp. 411-424.
Lee, J. S. IL, Haralic, R M. and Shapiro, L. G., 1987.
Morphologic edge detection. /EEE Journal of Robotics and
Automation, Vol. 3, No. 2, pp. 142-156.
Louverdis, G., Vardavoulia, M. I., Andreadis, I., and Tsalides,
Ph., 2002. A new approach to morphological color image
processing. Pattern Recognition, Vol. 35, pp. 1733 — 1741.
Mardia, K., 1976. A comment on the ordering of multivariate
data (Barnett, V.). Journal of Royal Statistical Society A,
Vol.3, pp. 318 - 355.
Mardia, K., Kent, J., and Bibby, J., 1979. Multivariate Analysis.
Academic Press, London.
Matheron, G., 1975. Random Sets and Integral Geometry. John
Wiley & Sons, New York.
Pratt, W. K., 1991. Digital Image Processing, John Wiley &
Sons, Interscience, New York.
Serra, J., 1982. Image Analysis and Mathematical Morphology.
Academic Press, London.
Talbot H, Evans C, Jones R, 1998. Complete ordering and
multivariate ^ mathematical morphology. Mathematical
Morphology and Its Applications to Image and Signal
Processing, Kluwer, pp. 27-34.
Titterington, D. M., 1978. Estimation of correlation coefficients
by ellipsoidal trimming. Applied Statistics, Vol. 27, pp. 277-
234.
Vardavoulia, M. I., Andreadis, L, Tsalides, P., 2002. Vector
ordering and morphological operations for color image
processing: fundamentals and applications. Pattern Analysis
and Applications, Vol. 5, pp. 271-287.
Winston, W. L., 1991. Introduction to Mathematical
Programming: Applications and Algorithms, Pws-Kent,
Boston.
