GS-SMC: A Camera Pose Refinement Framework for Monocular Visual Localization via 3-D Gaussian Splatting

Camera pose refinement aims to improve the accuracy of an initial estimation for camera position and orientation, ensuring reliable measurements in computer vision applications. Most refinement approaches rely on 2-D–3-D correspondences with specific descriptors or dedicated networks, requiring reconstructing the scene again for a different descriptor or fully retraining the network for each scene. Some recent methods instead infer pose from feature similarity, but their lack of geometry constraints results in less accuracy. To overcome these limitations, we propose a novel camera pose refinement framework leveraging 3-D Gaussian splatting (3DGS), referred to as GS-SMC. Given the widespread usage of 3DGS, our method can employ an existing 3DGS model to render novel views, providing a lightweight solution that can be directly applied to diverse scenarios without additional training or fine-tuning. Specifically, we introduce an iterative optimization approach, which refines the camera pose using epipolar geometric constraints (EGCs) among the query and multiple rendered images. Our method allows flexibly choosing feature extractors and matchers to establish these constraints. Extensive empirical evaluations on the 7-Scenes and the Cambridge Landmarks datasets demonstrate that our method significantly outperforms state-of-the-art camera pose refinement approaches, achieving median translation and rotation error reductions of 53.3% and 56.9% on 7-Scenes, and 40.7% and 51.6% on Cambridge.