{"slug": "gaussians-can-subtract-neural-dissolve-fields-for-hybrid-scene-rendering", "title": "Gaussians Can Subtract: Neural Dissolve Fields for Hybrid Scene Rendering", "summary": "A new arXiv paper (2611.07049v1) introduces Gaussolve, a subtract-and-refill hybrid rendering framework that selectively dissolves inconsistent mesh content before residual 3D Gaussians refill the revealed regions, rather than following the additive formulation common in existing Mesh-Gaussian scene reconstruction. The framework uses a lightweight Neural Dissolve Field, parameterized by scalar spherical-harmonic Gaussians, to learn where and from which viewing directions mesh content should be dissolved, paired with a continuous transmittance fusion rule that progressively exposes residual Gaussians while preserving their reconstruction gradients. Experiments on structured and semi-structured scenes show higher visual fidelity with fewer Gaussian primitives and faster rendering.", "body_md": "arXiv:2611.07049v1 Announce Type: new \nAbstract: Hybrid Mesh-Gaussian rendering combines compact textured surfaces with flexible 3D Gaussian primitives, offering an efficient alternative to purely Gaussian scene representations for structured environments. Existing scene-reconstruction hybrids commonly follow an additive formulation, where Gaussian residuals compensate for errors left by the mesh. However, not every Mesh-Gaussian conflict is caused by missing content: residual geometry errors, occlusion mismatch, texture-projection artifacts, and view-dependent appearance can make the same mesh content consistent in one view but inconsistent in another. We therefore formulate hybrid rendering as a multi-view inconsistency problem and present Gaussolve, a subtract-and-refill framework that selectively dissolves inconsistent mesh content before residual Gaussians refill the revealed regions. We introduce a lightweight Neural Dissolve Field, parameterized by scalar spherical-harmonic Gaussians, to learn where and from which viewing directions mesh content should be dissolved. A continuous transmittance fusion rule progressively exposes residual Gaussians behind the dissolved mesh, preserving their reconstruction gradients and avoiding hard visibility discontinuities. Experiments on structured and semi-structured scenes demonstrate higher visual fidelity with fewer Gaussian primitives and faster rendering. Project page: https://ooosn.github.io/gaussians-can-subtract/", "url": "https://wpnews.pro/news/gaussians-can-subtract-neural-dissolve-fields-for-hybrid-scene-rendering", "canonical_source": "https://arxiv.org/abs/2611.07049", "published_at": "2026-11-06 05:00:00+00:00", "updated_at": "2026-11-06 05:23:33.265193+00:00", "lang": "en", "topics": ["computer-vision", "neural-networks", "ai-research"], "entities": ["Gaussolve", "Neural Dissolve Field", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/gaussians-can-subtract-neural-dissolve-fields-for-hybrid-scene-rendering", "markdown": "https://wpnews.pro/news/gaussians-can-subtract-neural-dissolve-fields-for-hybrid-scene-rendering.md", "text": "https://wpnews.pro/news/gaussians-can-subtract-neural-dissolve-fields-for-hybrid-scene-rendering.txt", "jsonld": "https://wpnews.pro/news/gaussians-can-subtract-neural-dissolve-fields-for-hybrid-scene-rendering.jsonld"}}