ConeGS: Error-Guided Densification Using Pixel Cones for Improved Reconstruction with Fewer Primitives

2025-11-10

AI Summary

3D Gaussian Splatting (3DGS) achieves high-quality, real-time novel-view synthesis but suffers from suboptimal spatial distribution, redundant coverage, and inefficiency due to its clone-based densification strategy. To address this, we propose a pixel-cone error-driven densification framework: leveraging depth priors from Instant NGP, we construct observation-aligned pixel cones and actively insert geometry-agnostic Gaussians at predicted depth locations based on image-space reconstruction error. We further introduce a pre-activation opacity regularization and a dynamic primitive budget mechanism. Our method significantly improves reconstruction accuracy and rendering efficiency under sparse primitive budgets, consistently outperforming state-of-the-art approaches across diverse budget constraints—especially under resource-limited conditions—yielding a more compact and robust scene representation.

Technology Category

• Computer Vision: 3D Computer Vision
• Search and Optimization: Sampling/Simulation-based Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

3D Gaussian Splatting (3DGS) achieves state-of-the-art image quality and real-time performance in novel view synthesis but often suffers from a suboptimal spatial distribution of primitives. This issue stems from cloning-based densification, which propagates Gaussians along existing geometry, limiting exploration and requiring many primitives to adequately cover the scene. We present ConeGS, an image-space-informed densification framework that is independent of existing scene geometry state. ConeGS first creates a fast Instant Neural Graphics Primitives (iNGP) reconstruction as a geometric proxy to estimate per-pixel depth. During the subsequent 3DGS optimization, it identifies high-error pixels and inserts new Gaussians along the corresponding viewing cones at the predicted depth values, initializing their size according to the cone diameter. A pre-activation opacity penalty rapidly removes redundant Gaussians, while a primitive budgeting strategy controls the total number of primitives, either by a fixed budget or by adapting to scene complexity, ensuring high reconstruction quality. Experiments show that ConeGS consistently enhances reconstruction quality and rendering performance across Gaussian budgets, with especially strong gains under tight primitive constraints where efficient placement is crucial.

Problem

• Optimizes primitive distribution in 3D Gaussian Splatting
• Reduces required primitives while improving reconstruction quality
• Guides densification via image-space errors and geometric proxies

Innovation

• Pixel cones guide densification via error pixels
• Pre-activation opacity penalty removes redundant Gaussians
• Primitive budgeting adapts to scene complexity constraints