Certainty-guided Reasoning and Refinement Network for Camouflaged Object Detection

Camouflaged object detection (COD), which aims to segment objects that are highly similar to their background, is a valuable yet challenging task. Due to the interference of clutter and noise in the background, existing methods often struggle to avoid misleading and accurately segment the camouflaged object. In this paper, we propose a novel Certainty-guided Reasoning and Refinement Network (CRRNet) for COD. The core idea is to first explicitly reason under accurate knowledge of camouflaged objects, and then refine the uncertain regions by a certainty propagation strategy. To achieve this, we innovatively propose two modules: the Certainty-guided Reasoning Module and Uncertain Region Refinement Module. Experimental results prove that the proposed CRRNet outperforms 12 state-of-the-art methods on three COD benchmark datasets.