AD-GS: Alternating Densification for Sparse-Input 3D Gaussian Splatting

2025-09-13

AI Summary

Under sparse-view conditions, 3D Gaussian Splatting (3DGS) suffers from floating artifacts, geometric distortions, and overfitting. To address these issues, we propose an alternating densification framework: during high-density phases, Gaussians are rapidly spawned and optimized to enhance appearance fidelity; during low-density phases, aggressive opacity pruning and geometric regularization suppress noise and improve structural consistency. We further introduce two novel constraints—pseudo-view consistency and edge-aware depth smoothing—to jointly govern Gaussian density distribution and geometric quality. By periodically modulating model capacity growth, our method maintains training efficiency while significantly improving 3D reconstruction accuracy and rendering quality under sparse inputs. Extensive evaluations on multiple benchmark datasets demonstrate superior performance over state-of-the-art methods, effectively mitigating floating artifacts and geometric errors, and enabling more robust and fine-grained scene reconstruction.

Technology Category

• Computer Vision: 3D Computer Vision
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Search and Optimization: Non-convex Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

3D Gaussian Splatting (3DGS) has shown impressive results in real-time novel view synthesis. However, it often struggles under sparse-view settings, producing undesirable artifacts such as floaters, inaccurate geometry, and overfitting due to limited observations. We find that a key contributing factor is uncontrolled densification, where adding Gaussian primitives rapidly without guidance can harm geometry and cause artifacts. We propose AD-GS, a novel alternating densification framework that interleaves high and low densification phases. During high densification, the model densifies aggressively, followed by photometric loss based training to capture fine-grained scene details. Low densification then primarily involves aggressive opacity pruning of Gaussians followed by regularizing their geometry through pseudo-view consistency and edge-aware depth smoothness. This alternating approach helps reduce overfitting by carefully controlling model capacity growth while progressively refining the scene representation. Extensive experiments on challenging datasets demonstrate that AD-GS significantly improves rendering quality and geometric consistency compared to existing methods.

Problem

• Addresses sparse-view 3DGS artifacts like floaters
• Controls uncontrolled densification causing geometry inaccuracies
• Reduces overfitting through alternating densification and regularization

Innovation

• Alternating densification framework with controlled phases
• Photometric loss training for fine-grained details
• Opacity pruning with geometry regularization techniques