Image Processing Reference
In-Depth Information
variety of measurements, partly redundant and partly complementary. These measure-
ments may be punctual or time-integrated, bi-dimensional or instantaneous (images),
vertical profiles with time-integration or not, as well as three-dimensional information
(e.g. oceanic/atmospheric profiler/sounder at ground level, satellite-borne, ship-borne).
With the large amount of archives and numerical models representing the geophysical/
biological processes in mind, it is obvious that the quantity of information available to
describe and model the Earth and our environment rapidly increasing. Data fusion as
a subject is becoming increasingly relevant because it efficiently helps scientists to
extract precise and relevant knowledge from available information.
By itself, the process of data fusion is not new in environmental studies.
Meteorologists, for example, have been using it in weather prediction for decades.
In remote sensing, many of the classification procedures widely used in the field
are obviously relevant to data fusion. Data fusion allows the combination of the
measurements produced by classification, as well as the monitoring of the quality
of information generated by these measurements.
The European Association of Remote Sensing Laboratories (EARSeL) created a
“Special Interest Group for Data Fusion” in 1996 (Wald
2000
). This group has been contributing to a better under-
standing and use of data fusion in the field of Earth observa-
tion through organizing regular meetings for its members to
discuss and tackle the fundamentals of data fusion in remote
sensing. A series of bi-annual international conferences called
“Fusion of Earth Data: Merging point measurements, raster
maps and remotely sensed images” was launched in 1996 with
the aim of exploring this field of research and helping the
scientific community to fully understand the benefits of data
fusion in the Earth observation domain (Ranchin and Wald
1996a, 1998, 2000a
). A set of terms of reference emerged
from this work, including the definition of data fusion as “a
formal framework in which are expressed means and tools for
the alliance of data originating from different sources. It aims
at obtaining information of greater quality; the exact definition
of 'greater quality' will depend upon the application” (Wald
1999
). “Quality” is used in the definition as a generic word to
imply when the user is more satisfied by the results obtained
through a fusion process than without it.
Many techniques for data fusion already exist. The focus of this chapter is on the
fusion of images. The general approach in the fusion of images is to create a new
set of images, I, usually of reduced dimensions, from the original sets of images, as
the following equation indicated:
Data fusion: a
formal frame-
work in which
are expressed
means and tools
for the alliance of
data originating
from different
sources. It aims
at obtaining
information of
greater quality;
the exact
definition of
'greater quality'
will depend
upon the
application
(
)
I f A,B,C,D, …
(11.1)
where A, B, C, D, … are the original sets of images and eventually characteristics
derived.
These sets may or may not be commensurate, and could originate from various
modalities (e.g. panchromatic, microwave, hyperspectral) taken at different points
