In this paper, a novel procedure for extracting human settlement extent in high-resolution SAR images, based on local autocorrelation and morphological processing, is presented. The algorithm is based on two steps. Hints for human settlements by information fusion based on local indexes of spatial autocorrelation are the outputs of the first step. A morphological chain builds on top of them the final settlements' borders. Examples from very different areas, i.e., the Darfur region in Sudan and the Lombardy region in Italy, stress the robustness of the procedure as well as its effectiveness.