| (a) Initially we
show the result of attempting to classify single pixels. The
very broken pattern, caused by speckle noise, together with
a low probability of correct classification (31.9%),
indicate that some form of data averaging is essential. |
(b) Next we average
the complex polarised data over a 3x3 window before
classification. This clearly reduces the statistical error
caused by speckle leading to the improved classification
probability of 68.9%. |
