Introduction - Hyperspectral Image Fusion - page 3

Fig. 1.1 Illustration of the remote sensing process

for their representation. Thus, the coverage of a wide area is compromised by the
spatial resolution of images. The primary task of most of the remote sensing systems
is to classify the captured region into various segments. These segments, typically
known as land cover classes are decided by the application. Every pixel in the scene
is associated with one of the classes. However, when the spatial resolution of an
image is poor, two or more classes can be the part of a single pixel, and it may not be
always possible to correctly extract the details. The design and operational cost of
sensors with high spatial resolution is very high. Such sensors, when built, generate
a huge amount of data in order to cover a wide geographical region. One would like
to collect as much information as possible regarding the composition of the scene
for a better classification. The reflectance of the materials present in the scene varies
with the wavelength. Thus, the values of reflectance, if captured at different bands
of wavelengths, provide disparate information related to the scene contents. This
set of images can be obtained by collecting the response of the scene over multiple
spectral bands. A cost effective solution that does not compromise with either the
area coverage, or the spatial resolution, and yet provides a better information about
the scene is built in the form of spectral imaging.

1.1 Spectral Imaging

The remote sensing data get captured by the sensor for a given spatial resolution
at a particular wavelength. Various objects in the scene reflect, absorb, and emit
electromagnetic radiation depending upon their material composition. For a single