A New Segmentation Method for Images Degraded by Multiplicative Noise.
Matthias Soellner, Helmut Suess · elib (German Aerospace Center) · 1994
A Split-and-Merge Process adapted to the statistical characteristics of SAR images is presented. Starting with an initial partition the image is divided in homogeneous quadratic regions first. By hypothesistesting, similar neighbooring regions are found and merged iteratively afterwards. Some results of this algorithm with DLR-ESAR and JPL-AIR-SAR data are shown at last.