Image Processing Reference
In-Depth Information
digital number (DN) values (called the quantization level). The detector-related
effects are often characterized as the noise equivalent difference in radiance (NEDL)
or reflectance (NEDr). Since the recorded digital num-
ber is a discrete quantified measure of the originally
continuous radiance measurement (i.e., the energy
received at the sensor), the greater the quantization
level the closer the quantized image output approxi-
mates the original continuous data (Schowengerdt
1997
). Generally, quantization levels for most systems
range between 8-bits (256 gray levels) and 12-bits
(4,096 gray levels).
Successful analysis of remotely sensed imagery is dependent on sufficient radio-
metric resolution, meaning that the NEDL is sufficiently low and the quantization
level high enough to allow discrimination among the reflected or emitted energy
levels recorded for various phenomena of interest in urban environments. Conversely,
low radiometric resolution (such as 6-bit quantization level) can be prone to satura-
tion, impeding discrimination of urban environment features due to insufficient
contrast (Jensen and Cowen
1999
). The first space-borne imaging systems, such as
the Landsat MSS, used a 6-bit quantization level. Similarly, some of the earlier IRS
platform payloads output images at 7-bit quantization level, while other IRS series
record images at 8-bit and 16-bit quantization levels. The quantization level of most
imaging systems (Landsat TM and ETM+, SPOT, DMSP, IKONOS, ASTER, and
QuickBird) is generally between 8-bit and 12-bit.
radiometric resolution
refers to the
quantization level at
which analog radiance
measurements are
digitized and stored
as digital number
(DN) values
7.2.3
Spectral Resolution
Spectral resolution refers to the capability of an imaging system to measure and
record sensed radiation within discrete ranges of wavelengths (referred to as wave-
bands). The importance of spectral resolution in urban remote sensing analyses is
discussed in detail in Chapter 4 (see also Herold et al.
2004
). The historical trend
in the evolution of imaging systems has been toward increased spectral sensitivity
(i.e., increased number of bands and narrower wavelength ranges).
Imaging spectrometers or hyperspectral imagers are capable of sensing hun-
dreds of narrow wavebands across the UV, visible, NIR, SWIR, and TIR portions
of the spectrum (see Chapter 9 for a review of applications). The only civilian
spaceborne imaging spectrometer has been the Hyperion instrument on the NASA
EO-1 satellite. However, a number of airborne imaging spectrometers have and
continue to acquire hyperspectral imaging data. The key advantage of hyperspec-
tral image data (~10 nm spectral resolution) over broadband multispectral data
(~100 nm spectral resolution) is the ability to select specific and multiple narrow
wavebands that optimally discriminate urban surface materials or quantify urban
surface conditions.
