River Planform Extraction From High-Resolution SAR Images via Generalized Gamma Distribution Superpixel Classification

The extraction of river planforms from remotely sensed satellite images is a task crucial importance to many applications such as land planning, water resource monitoring, or flood prediction. In this article, we present novel framework for the rivers synthetic aperture radar (SAR) images, based on superpixel segmentation and subsequent classification. Superpixel achieved by modeling image pixels' amplitudes spatial coordinates finite mixture model, where generalized Gamma distribution used model accurately variety high-resolution SAR scenes. A number features describing texture statistics are extracted level, facilitating identification superpixels-planforms then unsupervised, agglomerative clustering, thus eliminating need labeled training data. We results our proposed method ICEYE-X2 SENTINEL-1 data, demonstrating its ability produce pixel-accurate masks.