3D Gabor Splatting: Reconstruction of High-frequency Surface Texture using Gabor Noise

2025-04-15

AI Summary

Standard 3D Gaussian splatting struggles to efficiently reconstruct high-frequency surface textures (e.g., fine stripes) because each anisotropic Gaussian kernel encodes only a single dominant color direction locally, necessitating excessive kernels to represent periodic details. To address this, we propose 3D Gabor splatting: a novel representation that embeds Gabor functions into Gaussian kernels, yielding 3D Gabor kernels that jointly possess local compact support and intrinsic periodic oscillation—enabling single-kernel modeling of multi-cycle textures. Within a differentiable splatting framework, we jointly optimize kernel density and Gabor texture parameters (frequency, orientation, phase). Experiments demonstrate substantial improvements in reconstruction fidelity on complex textured surfaces (e.g., fine stripes, wireframe patterns), achieving ~30% reduction in parameter count while recovering high-frequency details more completely. This approach overcomes the fundamental limitation of conventional Gaussian splatting in surface texture representation.

Technology Category

• Machine Learning: Kernel Methods
• Computer Vision: 3D Computer Vision
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text

Abstract

3D Gaussian splatting has experienced explosive popularity in the past few years in the field of novel view synthesis. The lightweight and differentiable representation of the radiance field using the Gaussian enables rapid and high-quality reconstruction and fast rendering. However, reconstructing objects with high-frequency surface textures (e.g., fine stripes) requires many skinny Gaussian kernels because each Gaussian represents only one color if viewed from one direction. Thus, reconstructing the stripes pattern, for example, requires Gaussians for at least the number of stripes. We present 3D Gabor splatting, which augments the Gaussian kernel to represent spatially high-frequency signals using Gabor noise. The Gabor kernel is a combination of a Gaussian term and spatially fluctuating wave functions, making it suitable for representing spatial high-frequency texture. We demonstrate that our 3D Gabor splatting can reconstruct various high-frequency textures on the objects.

Problem

• Reconstructing high-frequency surface textures in 3D models
• Reducing Gaussian kernel usage for fine texture representation
• Enhancing 3D texture detail with Gabor noise integration

Innovation

• 3D Gabor splatting for high-frequency textures
• Combines Gaussian and wave functions
• Enhances reconstruction of fine surface details