Structure-Guided Camouflaged Object Detection with Progressive Enhancement Strategy

Camouflaged Object Detection (COD) is a challenging task due to the inherent difficulty of distinguishing camouflaged objects from their highly similar backgrounds. Existing methods predominantly rely on structural cues but often suffer from misinterpretations and noise, especially when detecting small objects. To address these issues, we propose the Structure-Guided Network (SGNet), which progressively supplements structural information from points to regions. SGNet incorporates three key modules: the Key Point Local Enhancement (KLE) to enhance point-level detail, the Hybrid Resolution Adaptation (HRA) mechanism for integrating high-resolution features, and the Structure-Guided Patch (SGP) for selective high-resolution patch extraction based on object shape. Experimental results on three widely used COD datasets demonstrate that SGNet significantly outperforms state-of-the-art methods, achieving more accurate localization and finer edge segmentation, while minimizing background noise.