The International Archives of the Photo gramme try. Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
(a) Multispectral image
Figure 7: Road extraction example combining radiometric and
geometric features.
image can reach 1 Giga pixels often in several spectral bands.
In this situation, it is unreasonable to load the whole image in
the computer memory to process it. This is even truer for high
resolution imagery based mapping applications, where, in order
to cover the area of interest several satellite scenes are usually
used.
Thus, streaming techniques enable to read, process and write the
result progressively without having to load the entire image into
memory. Most of typical image processing operations work lo
cally and are compatible with streaming. Streaming capabilities
are included in OTB transparently for the user, thus enabling an
easy processing of huge images.
6.3 Efficient data processing
Since images are divided into separate streams and since most
current processors have two, four, or even more, cores, it is tempt
ing to use this capabilities process data in parallel and to save
some time. This property is called multithreading. Once again,
OTB automatically enables multithreading when it is possible.
This is especially valuable to work on clusters where the combi
nation of several processors help to greatly reduce the total pro
cessing time. These operations and the repartition on the different
computing units is transparent for the OTB user, thus removing
most of the burden of multithreaded programming.
7 CONCLUSION
In the mapping and photogrammetry fields the concepts used are
often difficult to handle and many users rely on commercial prod
ucts. However, to truly master the process, a hands on approach
is unavoidable.
In this context, OTB allows the user to practice and experiment
with real data and real tools and can be used in several situations:
• for engineers: efficient development on new remote sensing
applications;
• for engineers: quick evaluation of the performance of a par
ticular algorithm on a specific type of data;
• for researchers: fast prototyping of new algorithms;
• for researchers: library of existing algorithms to compare
the results with a new one;
• for professors: teaching of image processing;
• for students: benefit from many heavily-tested algorithms
popular in the literature.
The ORFEO Toolbox concerns all people working in the remote
sensing imagery community. Releasing it as an open source soft
ware, CNES hopes to benefit from the contribution of many spe
cialists to help grow the practical use of satellite imagery.
REFERENCES
Christophe, E. and Inglada, J., 2007. Robust road extraction for
high resolution satellite images. In: IEEE International Confer
ence on Image Processing, ICIP’07.
Inglada, J. and Giros, A., 2004. On the possibility of automatic
multi-sensor image registration. IEEE Trans. Geoscience and Re
mote Sensing.
1267