very bright, e.g. 255. In the example, digital number values of 49 and 106 are
assigned to 0 and 255, respectively. In the following contrast stretch the remainders
of the pixel values are distributed linearly between the assigned extreme values of
zero and 255. This results in an improved contrast ratio of most brightness values in
the original image. However, it should be noted that linear contrast stretch may also
result in a loss of contrast at the extreme DN values. This implies saturation, or
‘clipping’ of very low and very high DN values.
Another form of contrast enhancement is the non-linear contrast stretch,
whereby the intensity values of the original histogram are subject to a uniformly
distribution stretch, i.e. the original histogram is being redistributed to produce a
uniform population density of pixels along the DN axis. This enhancement is most
effective in providing better contrast for the most populated range of intensity
values in the original scene. The drawback of the uniform distribution stretch is
its compression of intensity values in the lower and upper ranges of the original
histogram, similar to the one encountered in the linear contrast stretch procedure.
However, a contrast stretch of the lower or upper ranges is required, and a
non-linear Gaussian stretch may be applied in order to accommodate this. The
Gaussian stretch fits the original components of the histogram to a normal distribution curve with zero and 255 as its lower and upper limit, respectively. The tradeoff is a relative loss of contrast in the moderate intensity range.
14.6 Simulation of SAR Imagery
Many aspects in the design of a SAR system have very large cost impacts.
Therefore, it is very important to fully understand the implications of specific
design and system parameters on the resultant imagery prior to the implementation
of a SAR. The question of how, for instance, a spaceborne SAR design can meet the
requirements of the user must be addressed. Some parameters can be modeled by
system engineers. However, the impact of other parameters, such as wavelength
and incidence angle, on the reflectivity of targets needs to be explored and evaluated for a particular application by means of image simulation.
By examining simulated imagery with known characteristics the system
designer and the user of SAR data alike obtain valuable information that helps to
optimize the system design. Furthermore, systematic image simulation can also
assist in the development of useful interpretation methodologies for spaceborne
SARs appropriate filtering techniques (Sabins 1987) for image enhancement, the
definition of SAR data compression and sampling techniques (Gibbins and Slaney
1991), as well as in the development of image and map registration techniques.
Consider this example from an applications point of view. By simulating a
spaceborne SAR with a 100 m 6-look resolution from a high resolution airborne
SAR image of sea ice in the Beaufort Sea some important image characteristics
could be derived. The simulations suggested that small hazards for shipping in these
ice infested waters, for example multiyear ice floes, pressure ridges and icebergs
14 Digital Processing of SAR Data and Image Analysis Techniques
293
assigned to 0 and 255, respectively. In the following contrast stretch the remainders
of the pixel values are distributed linearly between the assigned extreme values of
zero and 255. This results in an improved contrast ratio of most brightness values in
the original image. However, it should be noted that linear contrast stretch may also
result in a loss of contrast at the extreme DN values. This implies saturation, or
‘clipping’ of very low and very high DN values.
Another form of contrast enhancement is the non-linear contrast stretch,
whereby the intensity values of the original histogram are subject to a uniformly
distribution stretch, i.e. the original histogram is being redistributed to produce a
uniform population density of pixels along the DN axis. This enhancement is most
effective in providing better contrast for the most populated range of intensity
values in the original scene. The drawback of the uniform distribution stretch is
its compression of intensity values in the lower and upper ranges of the original
histogram, similar to the one encountered in the linear contrast stretch procedure.
However, a contrast stretch of the lower or upper ranges is required, and a
non-linear Gaussian stretch may be applied in order to accommodate this. The
Gaussian stretch fits the original components of the histogram to a normal distribution curve with zero and 255 as its lower and upper limit, respectively. The tradeoff is a relative loss of contrast in the moderate intensity range.
14.6 Simulation of SAR Imagery
Many aspects in the design of a SAR system have very large cost impacts.
Therefore, it is very important to fully understand the implications of specific
design and system parameters on the resultant imagery prior to the implementation
of a SAR. The question of how, for instance, a spaceborne SAR design can meet the
requirements of the user must be addressed. Some parameters can be modeled by
system engineers. However, the impact of other parameters, such as wavelength
and incidence angle, on the reflectivity of targets needs to be explored and evaluated for a particular application by means of image simulation.
By examining simulated imagery with known characteristics the system
designer and the user of SAR data alike obtain valuable information that helps to
optimize the system design. Furthermore, systematic image simulation can also
assist in the development of useful interpretation methodologies for spaceborne
SARs appropriate filtering techniques (Sabins 1987) for image enhancement, the
definition of SAR data compression and sampling techniques (Gibbins and Slaney
1991), as well as in the development of image and map registration techniques.
Consider this example from an applications point of view. By simulating a
spaceborne SAR with a 100 m 6-look resolution from a high resolution airborne
SAR image of sea ice in the Beaufort Sea some important image characteristics
could be derived. The simulations suggested that small hazards for shipping in these
ice infested waters, for example multiyear ice floes, pressure ridges and icebergs
14 Digital Processing of SAR Data and Image Analysis Techniques
293