Intrinsic decomposition and editing of 3D Gaussian splats

2026-06-30 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision & Graphics · Depth: Expert, quick

Summary

A new method extends intrinsic decomposition to 3D Gaussian splats, enabling advanced image editing capabilities for radiance fields. This approach models intrinsic decomposition using independent sets of Gaussian primitives, allowing each set to adapt to its specific layer characteristics. It also introduces an optimization procedure guided by data-driven predictions to disentangle multi-view photographs into these intrinsic sets. The proposed editing workflow allows users to modify the texture of planar surfaces by adjusting the albedo in a single image. This edit is then captured within the intrinsic radiance field, facilitating re-rendering of the modified scene with realistic lighting from any viewpoint. This technique builds on the long history of intrinsic decomposition in image editing, which separates object colors and textures from lighting.

Key takeaway

For 3D graphics artists or computer vision engineers developing scene editing tools, this intrinsic decomposition method for Gaussian splats offers a powerful new capability. You can now modify object textures, specifically planar surfaces, by simply adjusting their albedo in one image, and the changes will consistently apply across all viewpoints with plausible lighting. This streamlines workflows for creating editable 3D scenes from multi-view photographs, significantly enhancing creative control and efficiency in virtual content generation.

Key insights

Intrinsic decomposition for 3D Gaussian splats enables texture editing independent of lighting across viewpoints.

Principles

• Model decomposition with independent Gaussian primitive sets.
• Use data-driven optimization for disentanglement.
• Albedo edits propagate across viewpoints.

Method

An optimization procedure, guided by data-driven predictions, disentangles multi-view photographs into independent intrinsic Gaussian primitive sets, enabling texture modification and re-rendering.

In practice

• Modify planar surface textures via albedo editing.
• Re-render edited scenes with arbitrary viewpoints.

Topics

• 3D Gaussian Splatting
• Intrinsic Decomposition
• Radiance Fields
• Image Editing
• Texture Editing
• Computer Graphics

Best for: Research Scientist, AI Scientist, Computer Vision Engineer

Related on AIssential

• Intrinsic 4D Gaussian Segmentation from Scene Cues — Intrinsic-GS segments 4D Gaussian scenes by leveraging inherent Gaussian properties and graph partitioning, eliminating external mask dependencies.
• Fi-Gaussian: Frequency-Aware Implicit Gaussian Splatting for Single Image Dehazing — Fi-Gaussian uses frequency-aware implicit Gaussian splatting and physics-driven renormalization for state-of-the-art single image dehazing.
• What happens when you combine Nano Banana with Gaussian splat tours? — Combining Gaussian splats with generative AI enables interactive spatial reimagining of real-world 3D environments.
• Less Gaussians, Texture More: 4K Feed-Forward Textured Splatting — LGTM enables 4K novel view synthesis by decoupling geometry from resolution using textured Gaussian primitives.
• FLUX3D: High-Fidelity 3D Gaussian Generation with Diffusion-Aligned Sparse Representation — FLUX3D enhances 3D Gaussian Splatting generation fidelity by resolving representation and cross-modal alignment bottlenecks.
• WildSplatter: Feed-forward 3D Gaussian Splatting with Appearance Control from Unconstrained Images — WildSplatter enables fast, pose-free 3DGS from unconstrained images by disentangling geometry and appearance via learned embeddings.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.