Stereoscopic viewing reveals hidden flaws in 3d gaussian splatting for vr
Beyond Monoscopic Viewing: A Study on 3D Gaussian Splatting Quality in VR
Human-Computer Interaction
Summary
3D Gaussian Splatting is a recent technology that creates 3D scenes from limited photos for virtual reality (VR). The authors found that when viewing these 3D scenes in VR with both eyes (stereoscopically), some image errors become very obvious, especially where objects appear to float or are in the wrong depth. Surprisingly, common methods used to measure image quality don’t catch these issues well, and looking with one eye (monoscopically) also hides them. Their study shows that testing 3D scene quality for VR needs to be done with both eyes to accurately spot problems.
What this means in practice
- •For vr content creators: identify and fix depth-related artifacts in 3D Gaussian Splatting scenes when preparing VR environments to improve viewer experience.
- •For quality assurance teams: develop stereoscopic-based tests to better evaluate 3D scene reconstructions for VR products, ensuring perceived visual quality matches user expectations.
Authors
Shreyas Shivakumara, Gabriel Eilertsen, Karljohan Lundin Palmerius
Abstract
Stereoscopy is fundamental to virtual reality (VR), providing depth perception through binocular viewing. Recent advances in 3D Gaussian Splatting (3DGS) enable high-quality novel view synthesis, making it well suited to immersive VR. We render 3DGS reconstructions stereoscopically and evaluate them in a head-mounted display, replicating how they would actually be viewed in VR. Real-world capture provides only a limited number of views, and under this constraint 3DGS reconstruction often produces localized floaters and misplaced structures. Standard metrics miss these localized artifacts, which become salient under stereoscopic viewing, where geometry is placed at the wrong depth. We investigate whether standard image-quality evaluation reflects the perceptual quality of 3DGS reconstructions under reduced capture. We compare SfM-only baseline with a union initialization that combines SfM with a dense VGGT network. All other training components are held fixed, isolating the effect of initialization coverage. We conduct a user study comparing preferences under monoscopic and stereoscopic HMD viewing, and test whether image-quality metrics predict the observed preferences. Monoscopically, preference for the more consistent reconstruction is weak, reaching 58.4\% overall. Stereoscopically, the same preference rises to 78.2\% and is consistent across all participants, while image-quality metrics (PSNR, SSIM and LPIPS) and stereo-aware metrics (iSQoe and StereoQA) show only modest differences and fail to penalize them. Our results indicate that monoscopic evaluation and standard image-quality metrics substantially underestimate perceptual artifacts observed in 3DGS reconstructions for VR, making stereoscopic assessment essential for 3DGS quality evaluation in VR.