RUN: Reversible Unfolding Network for Concealed Object Segmentation

Concealed object segmentation (COS) is a challenging problem that focuses on identifying objects that are visually blended into their background. Existing methods often employ reversible strategies to concentrate on uncertain regions but only focus on the mask level, overlooking the valuable of the RGB domain. To address this, we propose a Reversible Unfolding Network (RUN) in this paper. RUN formulates the COS task as a foreground-background separation process and incorporates an extra residual sparsity constraint to minimize segmentation uncertainties. The optimization solution of the proposed model is unfolded into a multistage network, allowing the original fixed parameters to become learnable. Each stage of RUN consists of two reversible modules: the Segmentation-Oriented Foreground Separation (SOFS) module and the Reconstruction-Oriented Background Extraction (ROBE) module. SOFS applies the reversible strategy at the mask level and introduces Reversible State Space to capture non-local information. ROBE extends this to the RGB domain, employing a reconstruction network to address conflicting foreground and background regions identified as distortion-prone areas, which arise from their separate estimation by independent modules. As the stages progress, RUN gradually facilitates reversible modeling of foreground and background in both the mask and RGB domains, reducing false-positive and false-negative regions. Extensive experiments demonstrate the superior performance of RUN and underscore the promise of unfolding-based frameworks for COS and other high-level vision tasks. Code is available at https://github.com/ChunmingHe/RUN.

This paper tackles the problem of finding hidden objects in images—objects that blend into the background and are hard to detect, even for humans. Most current methods try to fix this by focusing only on certain areas of the object mask, ignoring helpful information in the actual image colors. To improve this, we introduce a new method called RUN, which separates the image into foreground (the object) and background in a step-by-step way. Our method looks at both the object mask and the image itself to better tell the difference between what should be kept and what should be ignored. It does this by reusing information and gradually improving its guesses in each step. As a result, it makes fewer mistakes and finds hidden objects more accurately. We tested our method on several datasets, and it performed better than other approaches. We plan to release the code to the public.