Neural Shell Texture Splatting: More Details and Fewer Primitives

2025-07-27

AI Summary

Gaussian splatting achieves high fidelity and rendering efficiency but suffers from strong coupling between geometry and appearance, necessitating a large number of Gaussian primitives for high-quality reconstruction. To address this, we propose Neural Shell Texture—a novel representation that decouples geometric and appearance modeling: explicit Gaussian primitives encode the scene geometry and serve as anchor points for texture sampling, while a learnable shell-shaped texture field enables global, continuous surface appearance modeling. This design supports differentiable rendering and efficient image-space projection. It significantly reduces the required number of primitives (by over 60% on average), enhances texture detail recovery, and enables high-fidelity mesh extraction. Experiments across multiple datasets demonstrate that our method outperforms state-of-the-art approaches, achieving a superior trade-off between reconstruction fidelity and rendering efficiency.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Representation Learning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Gaussian splatting techniques have shown promising results in novel view synthesis, achieving high fidelity and efficiency. However, their high reconstruction quality comes at the cost of requiring a large number of primitives. We identify this issue as stemming from the entanglement of geometry and appearance in Gaussian Splatting. To address this, we introduce a neural shell texture, a global representation that encodes texture information around the surface. We use Gaussian primitives as both a geometric representation and texture field samplers, efficiently splatting texture features into image space. Our evaluation demonstrates that this disentanglement enables high parameter efficiency, fine texture detail reconstruction, and easy textured mesh extraction, all while using significantly fewer primitives.

Problem

• Reduces primitive count in Gaussian splatting techniques
• Disentangles geometry and appearance for better efficiency
• Enhances texture detail with neural shell representation

Innovation

• Neural shell texture encodes global surface texture
• Gaussian primitives as geometry and texture samplers
• Disentangled geometry and appearance for fewer primitives