ABSTRACT
Sea ice synthetic aperture data (SAR) data are an important source of navigational information in the ice-covered seas, one of these areas being the northern parts of the Baltic Sea, especially Gulf of Finland with a very heavy winter sea traf†c. To automate the interpretation of SAR images, several sea ice classi†cation algorithms have been developed at the Finnish Meteorological Institute. To improve our classi†cation schemes, more detailed information on sea ice types is required. One way to better utilize the information present in sea ice SAR data is to better utilize the textural information. We have studied the use of the independent component analysis [1] to extract basic textural primitives from SAR data to be used in classi†cation [2]. To improve this feature extraction, we have also studied some novel methods to extract some elementary texture features from sea ice SAR data.
