GOR-IS: 3D Gaussian Object Removal in the Intrinsic Space

2026-05-01

AI Summary

Existing 3D Gaussian splatting–based object removal methods struggle to handle non-Lambertian surfaces and global illumination consistency, often yielding physically implausible and visually incoherent inpainting results. This work addresses these limitations by introducing intrinsic image decomposition and explicit light transport modeling into the 3D Gaussian splatting framework for the first time. We propose an intrinsic-space inpainting module that performs edits at the level of intrinsic image components, thereby enforcing physical consistency. Our approach effectively tackles view-dependent reflectance and light propagation challenges, achieving significantly improved performance over state-of-the-art methods on both synthetic and real-world datasets—demonstrated by a 13% reduction in LPIPS and a 2 dB gain in PSNR—producing more realistic and temporally coherent object removal results.

Abstract

Recent advances in Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) have made it standard practice to reconstruct 3D scenes from multi-view images. Removing objects from such 3D representations is a fundamental editing task that requires complete and seamless inpainting of occluded regions, ensuring consistency in geometry and appearance. Although existing methods have made notable progress in improving inpainting consistency, they often neglect global lighting effects, leading to physically implausible results. Moreover, these methods struggle with view-dependent non-Lambertian surfaces, where appearance varies across viewpoints, leading to unreliable inpainting. In this paper, we present 3D Gaussian Object Removal in the Intrinsic Space (GOR-IS), a novel framework for physically consistent and visually coherent 3D object removal. Our approach decomposes the scene into intrinsic components and explicitly models light transport to maintain global lighting effects consistency. Furthermore, we introduce an intrinsic-space inpainting module that operates directly in the material and lighting domains, effectively addressing the challenges posed by non-Lambertian surfaces. Extensive experiments on both synthetic and real-world datasets demonstrate that our framework substantially improves the physical consistency and visual coherence of object removal, outperforming existing methods by 13% in perceptual similarity (LPIPS) and 2dB in peak signal-to-noise ratio (PSNR). Code is publicly available at https://applezyh.github.io/GOR-IS-project-page/

Problem

• 3D object removal
• inpainting consistency
• global lighting effects
• non-Lambertian surfaces
• 3D Gaussian Splatting

Innovation

• Intrinsic Decomposition
• Light Transport Modeling
• Non-Lambertian Surfaces
• 3D Gaussian Splatting
• Physically Consistent Inpainting