Classification of remote-sensing images by using the Bayes rule for minimum cost

An approach based on the Bayes rule for minimum cost for feature selection and classification of remote-sensing images is proposed. This approach allows one to achieve land-cover maps in which the total cost involved by errors, instead of the total classification error, is minimized. Experiments carried out on a multisource data set of the Island of Elba (Italy) point out the effectiveness of the proposed minimum cost approach.