this we can distinguish edges created by geometric and ra-
diometric discontinuities in the object space, by comparing
the scale space of the image to the scale spaces of the object
space.
5. COMMENTS
Scale space can be used to represent two dimensional sig-
nals in various resolutions. This representation can thus be
used for images as well as for radiometric and/or geomet-
ric descriptions of the object space. It is structured and
explorable and it can offer valuable assistance in various
photogrammetric processes.
The concept of scale space provides the theoretical founda-
tion for hierarchical implementation of digital photogram-
metric tasks, allowing otherwise cumbersome and time con-
suming modules to be performed quickly and effectively. For
instance, automatic stereopair orientation can be performed
using digital image pyramids to effectively lead the results
to continuously improving accuracies [Schenk et al., 1991].
However, besides implementing some modules in a hierar-
chical fashion, scale space theory can also be used to refine
the performance of well-established processes, such as least
squares matching and orthophoto production. By investi-
gating the differential scale variations which exist between
conjugate features in different images, we can deduce a scale-
adapting matching process aiming at the optimization of
least squares matching. In orthophoto production, we can
bring features to the same radiometric and geometric level
of scale space, thus eliminating discrepancies and improving
its overall performance.
In general, the advantage of using scale space theory to rep-
resent the object space is twofold. Signals describing the
object space can be stored in a compact yet efficient way
by recording their discontinuities through scale space and
in addition, image and object space can be directly com-
pared and semantic information can be extracted from this
comparison.
References
[1] Agouris P., Schenk A.F. & Stefanidis A. (1989) Zero-
Crossings for Edge Detection, Proceedings 1989 ASPRS
Fall Convention, Cleveland, pp. 91-99.
[2] Babaud J., Witkin A., Baudin M. & Duda R.O. (1986)
Uniqueness of the Gaussian Kernel for Scale-Space Fil-
tering, IEEE Transactions on Pattern Analysis and Ma-
chine Intelligence, Vol. 8, No. 1, pp. 26-33.
[3] Bergholm F. (1987) Edge Focusing, IEEE Transactions
on Pattern Analysis and Machine Intelligence, Vol. 9,
No. 6, pp.726-741.
[4] Chakreyavanich U. (1991) Regular DEM Data Com-
pression by Using Zero Crossings: The Automatic
Breakline Detection Method, Report No. 412, Dept. of
Geodetic Science, The Ohio State University.
[5] Doorn B.D. (1991) Multi-Scale Surface Reconstruction
in the Object Space, Report No. 413, Dept. of Geodetic
Science, The Ohio State University.
590
[6] Lindeberg T. (1990) Scale-Space for Discrete Signals,
IEEE Transactions on Pattern Analysis and Machine
Intelligence, Vol. 12, No. 3, pp. 234-254.
[7] Lu Y. & Jain R.C. (1989) Behavior of Edges in Scale
Space, IEEE Transactions on Pattern Analysis and Ma-
chine Intelligence, Vol. 11, No. 4, pp. 337-356.
[8] Novak K. (1992) Rectification of Digital Imagery, Pho-
togrammetric Engineering & Remote Sensing, Vol. 58,
No. 3, pp. 339-344.
[9] Schenk A.F., Li J.C. & Toth C. (1991) Towards an
Autonomous System for Orienting Digital Stereopairs,
Photogrammetric Engineering & Remote Sensing, Vol.
57, No. 8, pp. 1057-1064.
[10] Weaver J.H. (1983) Applications of Discrete and Con-
tinuous Fourier Analysis, John Wiley & Sons.
[11] Witkin A.P. (1983) Scale-Space Filtering, Proceedings
7th International Joint Conference on Artificial Intelli-
gence, Karlsruhe, pp. 1019-1022.
sr
[12] Yuille A.L. & Poggio T. (1983) Fingerprints Theorems
for Zero-Crossings, A.I. Memo 730, Artificial Intelli-
gence Laboratory, Massachusetts Institute of Technol-
ogy.
