sub-beams. Each of these sub-beams, or sub-bands, provides an independent ‘look’
at the illuminated scene. This results in an enlarged resolution cell length and a
number of independent images. The distribution of speckle levels within each
‘look’ will also be independent of each other. If the ‘looks’ are summed, resulting
image intensity will have reduced speckle components. In the case where N
independent images have been formed and systematically averaged the speckle
variance is reduced by a factor of N.
The maximum number of “looks”, i.e. non-overlapped sub-apertures, over the
full azimuth beam width is seldom processed. There are two reasons: to minimize
signal ambiguities, and to maintain a reasonable image resolution. The Latter point
is of some consequence. The trade-off in Multi-look processing is that the available
resolution of the image is also degraded by a factor of N, because in order to
generate independent ‘looks’, different portions of the original signal must be used.
Consider this example: the azimuth bandwidth of the SEASAT SAR data is
sufficient to allow for a maximum azimuth resolution of approximately 6.5 m, if
the signal is processed to full coherence; multi-look processing, and speckle
reduction as a result, achieved 25 m image resolution by taking four independent
“looks”.
The number of ‘looks’ one may desire depends very much on the SAR application field in mind and may vary accordingly. Those applications concerned with
fine detailed structure or small area targets may place greater emphasis on high
spatial resolution; other applications which require good radiometric resolution
may choose a larger number of ‘looks’ when specifying SAR processing options.
It is necessary to specify both the spatial resolution and associated number of lookrequirements.
14.3 Radiometric Correction of Digital SAR Data
The radiometric fidelity of SAR imagery is affected by intensity variations resulting
from surface scattering geometry and antenna pattern variations (Sabins 1987). The
surface scattering geometry causes radiometric distortions, because at increasing
incidence angle down-range less power is received. This causes less intense signal
returns and less image brightness. Reference is made to the ‘radar equation’
(Eq. 14.1), which states that the power received is inversely proportional to the
fourth power of the range. This relationship is known as the R
4 power loss. The
antenna pattern causes radiometric distortion in the range dimension, because an
antenna transmits more power from the centre of the antenna than from its edges.
This results in more intense radar returns in the mid-range of the image swath
relative to the near- and far-range edges where illumination is less intense.
Simple empirical techniques for correcting radiometric distortions in a SAR
image (Richards and Jia 1999) may include the following steps. At various range
locations, areas with the same surface state are identified. An average intensity
value of the pixels is computed for each area and plotted versus range. Then a
14 Digital Processing of SAR Data and Image Analysis Techniques
285
at the illuminated scene. This results in an enlarged resolution cell length and a
number of independent images. The distribution of speckle levels within each
‘look’ will also be independent of each other. If the ‘looks’ are summed, resulting
image intensity will have reduced speckle components. In the case where N
independent images have been formed and systematically averaged the speckle
variance is reduced by a factor of N.
The maximum number of “looks”, i.e. non-overlapped sub-apertures, over the
full azimuth beam width is seldom processed. There are two reasons: to minimize
signal ambiguities, and to maintain a reasonable image resolution. The Latter point
is of some consequence. The trade-off in Multi-look processing is that the available
resolution of the image is also degraded by a factor of N, because in order to
generate independent ‘looks’, different portions of the original signal must be used.
Consider this example: the azimuth bandwidth of the SEASAT SAR data is
sufficient to allow for a maximum azimuth resolution of approximately 6.5 m, if
the signal is processed to full coherence; multi-look processing, and speckle
reduction as a result, achieved 25 m image resolution by taking four independent
“looks”.
The number of ‘looks’ one may desire depends very much on the SAR application field in mind and may vary accordingly. Those applications concerned with
fine detailed structure or small area targets may place greater emphasis on high
spatial resolution; other applications which require good radiometric resolution
may choose a larger number of ‘looks’ when specifying SAR processing options.
It is necessary to specify both the spatial resolution and associated number of lookrequirements.
14.3 Radiometric Correction of Digital SAR Data
The radiometric fidelity of SAR imagery is affected by intensity variations resulting
from surface scattering geometry and antenna pattern variations (Sabins 1987). The
surface scattering geometry causes radiometric distortions, because at increasing
incidence angle down-range less power is received. This causes less intense signal
returns and less image brightness. Reference is made to the ‘radar equation’
(Eq. 14.1), which states that the power received is inversely proportional to the
fourth power of the range. This relationship is known as the R
4 power loss. The
antenna pattern causes radiometric distortion in the range dimension, because an
antenna transmits more power from the centre of the antenna than from its edges.
This results in more intense radar returns in the mid-range of the image swath
relative to the near- and far-range edges where illumination is less intense.
Simple empirical techniques for correcting radiometric distortions in a SAR
image (Richards and Jia 1999) may include the following steps. At various range
locations, areas with the same surface state are identified. An average intensity
value of the pixels is computed for each area and plotted versus range. Then a
14 Digital Processing of SAR Data and Image Analysis Techniques
285