Introduction - Hyperspectral Image Fusion - page 8

Table 1.2 Specifications of Hyperion system
Year of availability 2002

Number of bands 220

Spectral range 0 . 4-2 . 5 µ m

Bandwidth at FWHM 10 nm

Spatial resolution 30 m

Scan line width 7.5 km

Source http://eo1.gsfc.nasa.gov

The Hyperion was the first spaceborne hyperspectral sensor to acquire the
reflectance from near IR and short wave infrared (SWIR) band [77]. The imagery is a
part of EO-1 spacecraft launched by NASA. Hyperion provides 10-bit data covering
a bandwidth of 0

m in 220 bands. This instrument can cover a 7.5 km
by 100 km region per image. Table 1.2 summarizes the basic specifications of the
Hyperion.
The radiance acquired by the sensor cannot be directly used for further operations
due to illumination and atmospheric effects. These effects mainly include the angle of
the sun, viewing angle of the sensor, the changes in solar radiance due to atmospheric
scattering [160, 165]. It also suffers from geometric distortions caused due to the
earth's curvature. Further operations are made possible after atmospheric corrections
which transform radiance spectra into reflectance. More details on the corrections
can be found in [29, 160, 168]. The raw imagery obtained from the sensors is
referred to as the Level 0 data. At the next level of processing, called Level 1R, the
data are calibrated to physical radiometric units, and formatted into image bands.
Later, the data are geometrically corrected, compensated for atmospheric effects,
and localized on the earth. These data are referred to as the Level 1G. One of the
important requirements for the processing and better understanding of these data is
to have all the bands of the hyperspectral image depicting exactly the same region. In
other words, all the bands should be spatially aligned to each other. This process is
known as image registration. We deal with only the co-georegistered Level 1G data
in this monograph.
Figure 1.3 illustrates the 3-Dnature of the hyperspectral data. It is often convenient
to visualize the hyperspectral image as a 3-D cube. One may observe a set of 2-D
bands that can be thought as the “slicing” of the cube along the spectral axis. When
the spectral signature is to be analyzed, it can be accomplished by observing an
1-D array of intensity values across all the bands for the specific spatial location
under investigation. The left portion of the figure depicts one of the bands of the
hyperspectral data. Observations along the wavelength axis at any spatial location
form an array of length equal to the number of bands. Two such representative arrays
are shown in the plot. These plots refer to the spectral signatures of the material
compositions of the corresponding pixels.
The 3-dimensional structure of the hyperspectral images indicates three different
structures for the storage of the data. If the hyperspectral image is to be primar-
ily regarded as the set of independent bands, it is beneficial to store it band-wise.

. 4to2 . 5 µ