Abstract: Highlights•We propose an Edge and Texture Awareness Module that refines segmentation boundaries.•We design a Prior Enhancement Module to refine features and enhance prior knowledge integration.•We propose a Hierarchical Fusion Module with four-layer attention for noise filtering and feature fusion.•Experiments on five datasets show SGNet outperforms state-of-the-art methods.
External IDs:dblp:journals/eswa/ChenXFSW25
Loading