The automatic interpretation of the high-resolution SAR sea
ice images is a difficult problem due to the complex interaction
between the ice or water surface and the SAR signal. The
magnitude of the reflected SAR signal is largely determined by
the radar frequency, the radar incidence angle, and the target
conditions such as surface roughness, moisture, snow cover,
and salinity [6]. For example, calm water generally has low
backscatter compared with wind-roughened water, most noticeably
in horizontal transmit and horizontal receive (HH) pol
and vertical transmit and vertical receive (VV) pol SAR images
compared with horizontal transmit and vertical receive (HV)
pol [6]. This is not always the case due the effect of the
incidence angle. The sensitivity of scattered SAR signals to
wind speed varies with incidence angle. The appearance of ice
in SAR images also changes through the ice season due to
melting, ice dynamics, and floe distribution. These factors make
it challenging to design image features to capture the diverse
patterns of ice and water in SAR imagery.