In this paper, a supervised CNN model is used as a feature
extractor and a regression model to estimate ice concentration
from SAR images. The processing scheme is shown in Fig. 3.
It is composed of three major steps: preprocessing of the SAR
images, training of the CNN model, and prediction using the
trained CNN model. Performance of the trained CNN model is
evaluated by testing on independent images to the images used
in training. Prediction follows the same procedure as testing,
except that an ice concentration is estimated for every pixel
location in the input SAR image; therefore, only the first two
steps will be described.