Image Processing Reference
are very similar to those in the Pan image. However, a close examination reveals
differences: the gray values and local contrasts are not the same as exemplified in
the upper left corner of the image or the above-mentioned large buildings. The gray
values in the synthesized image are very close to those in the original NIR image.
In summary, the ARSIS concept provides an excellent framework for the devel-
opment of accurate methods that can be tailored to specific user needs and can be
assembled within a toolbox. It is also an open framework with many areas for the
development of different cases of applications and approaches for implementation.
Different methods can be developed based on this concept, depending upon the
multiscale description and synthesis model MSM, the IBSM relating the content of
both representations A and B , and the HRIBSM transforming the parameters of the
IBSM when increasing the spatial resolution (Wald 2002 ; Ranchin and Wald
2000b ; Ranchin et al. 2003 ).
The methods belonging to the ARSIS concept are constructed in such a way that
once the synthesized images B* h are degraded to a lower resolution l , they reproduce
the original spectral content of B l . These methods perform a high-quality transforma-
tion of the multispectral content when increasing the spatial resolution. The synthe-
sized images can therefore be used for purposes other than visual interpretation.
In the case of the SPOT imagery, it was demonstrated that the accuracy and the quality
of the road network automatically extracted was increased by the use of images
synthesized by the ARSIS concept (Ranchin and Wald 1997 ). For urban mapping,
the benefits of using ARSIS synthesized images were enhanced by Ranchin and
Wald ( 1996b ), Raptis et al. ( 1998 ), Terretaz ( 1997 ), and Couloigner et al. ( 1998 ).
The benefits of ARSIS fused products for the analysis and mapping of the city were
demonstrated over the city of Marseille (France) through the use of color images
having a higher spatial resolution clearly permits a more accurate interpretation of
the features in the city (Wald and Ranchin 2001 ). Furthermore, such fused products
may be the object of further image processing techniques, such as contrast enhance-
ment, edge detection or classification, without the creation of visible artifacts.
It illustrates the capability of the synthetic images to support further image process-
ing dealing with the high frequencies (Wald and Ranchin 2001 ). The ARSIS concept
was also employed to increase the quality in mapping air quality in city (Wald and
Baleynaud 1999 ).
Fusion of Images, Databases and Punctual Measurements
for Air Quality
Presently, most large cities in Europe have an air quality surveillance program.
Such a network is composed of a few static measuring stations, which perform a
continuous surveillance of air pollution at station locations. In France, pollution