RevealLayer: Disentangling Hidden and Visible Layers via Occlusion-Aware Image Decomposition

Recent diffusion-based approaches have made substantial progress in image layer decomposition. However, accurately decomposing complex natural images remains challenging due to difficulties in occlusion completion, robust layer disentanglement, and precise foreground boundaries. Moreover, the scarcity of high-quality multi-layer natural image datasets limits advancement. To address these challenges, we propose RevealLayer, a diffusion-based framework that decomposes an RGB image into multiple RGBA layers, enabling precise layer separation and reliable recovery of occluded content in natural images. RevealLayer incorporates three key components: (1) a Region-Aware Attention module to disentangle hidden and visible layers; (2) an Occlusion-Guided Adapter to leverage contextual information to enhance overlapping regions; and (3) a composite loss to enforce sharp alpha boundaries and suppress residual artifacts. To support training and evaluation, we introduce RevealLayer-100K, a high-quality multi-layer natural image constructed through a collaboration between automated algorithms and human annotation, and further establish RevealLayerBench for benchmarking layer decomposition in general natural scenes. Extensive experiments demonstrate that RevealLayer consistently outperforms existing approaches in layer decomposition.

Many real-world images contain overlapping objects, where some parts of the scene are hidden behind others. Separating such an image into clean, editable layers is useful for image editing, object removal, and scene understanding, but it remains difficult because a system must decide which pixels belong to each object and also infer what the hidden regions should look like. We present RevealLayer, a system that takes a single natural image and separates it into multiple transparent layers. These layers can represent the background and different foreground objects, making it easier to edit or understand the scene. RevealLayer is designed to produce clear object boundaries, reduce unwanted artifacts, and recover missing content in regions where objects overlap. We also introduce a large high-quality dataset of layered natural images, created with both automated tools and human annotation, as well as a benchmark for evaluating this task. Our experiments show that RevealLayer performs better than existing methods at producing clean and complete image layers in complex natural scenes.