StereoGS: Sparse-View 3D Gaussian Splatting via Stereo Priors

2026-06-29

AI Summary

This work addresses the geometric distortion in 3D Gaussian Splatting under sparse-view settings, where the absence of geometric constraints leads to overfitting. To mitigate this, we introduce stereo vision priors into the framework for the first time by constructing virtual stereo pairs and enforcing absolute-scale and binocular consistency constraints during optimization. Additionally, we propose a gradient-aware opacity decay mechanism and a consistency-aware dense initialization strategy. These components collectively alleviate monocular scale ambiguity and cross-view inconsistency without incurring additional inference overhead. Our method achieves state-of-the-art reconstruction quality on benchmark datasets—including LLFF, DTU, Mip-NeRF360, and Blender—under sparse input conditions.

Abstract

3D Gaussian Splatting (3DGS) has achieved remarkable success in real-time novel view synthesis, yet it suffers from severe overfitting under sparse-view settings due to insufficient geometric constraints. While recent methods introduce monocular depth priors to mitigate this, they inherently struggle with scale ambiguity and cross-view inconsistency, leading to defective geometry. In this paper, we propose StereoGS, a novel sparse-view 3DGS framework that integrates stereo priors to establish reliable binocular consistency. Unlike scale-agnostic monocular constraints, StereoGS introduces a Stereo Depth Regularization by constructing virtual stereo pairs during optimization and leveraging a foundation stereo model to enforce absolute scale and binocular-consistent structures. To further suppress overfitting and eliminate redundant primitives, we design a Gradient-Aware Opacity Decay strategy that dynamically penalizes Gaussians based on their relative opacity gradient magnitudes. Combined with a Consistency-Aware Dense Initialization using zero-shot multi-view depth estimation, StereoGS effectively anchors primitives to accurate scene surfaces. Extensive experiments on LLFF, DTU, Mip-NeRF360, and Blender datasets demonstrate that StereoGS achieves state-of-the-art performance in sparse-view settings without incurring any additional inference overhead. Project Page: https://stringerywh00.github.io/StereoGS_project_page/

Problem

• 3D Gaussian Splatting
• sparse-view
• overfitting
• geometric constraints
• novel view synthesis

Innovation

• Stereo Priors
• 3D Gaussian Splatting
• Sparse-View Reconstruction
• Depth Regularization
• Opacity Decay