Map generalization: Information theoretic approach to feature elimination

The present paper proposes an entropy based algorithm for the feature elimination in cartographic generalization. The methodology is tested on a map with 100 point symbols. The experiment demonstrates how the most con°icting map symbols are detected and eliminated from the map. The optimal symbol size can be computed from the entropy model. It is shown how to modify the entropy measure in such a way that the point elimination to a high degree preserves the point clusters.