D^2-4DGS: Dual-Depth Guided Sparse-Camera 4D Gaussian Splatting
2026-08-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors present a method called D²-4DGS to create 3D video-like scenes from a few camera views, which is usually hard because fewer cameras mean less detailed shape information. They combine two types of depth information: one from single cameras that gives lots of detail but can be imprecise in scale, and another from multiple cameras that is accurate but incomplete. By aligning and checking these depth sources, they keep the good parts and fix missing areas, making the 3D scenes look better and more consistent. Their approach improves the quality of generated views compared to previous methods using fewer cameras.
4D Gaussian Splattingdynamic novel view synthesissparse multi-view videosmonocular depth estimationmulti-view geometric depthdepth priorsRGB-D joint optimizationPSNRscene reconstructiondepth supervision
Authors
Jijian Zhao
Abstract
Dynamic 4D Gaussian Splatting has emerged as an efficient representation for dynamic novel view synthesis through explicit scene modeling and real-time rendering. However, existing methods typically require dense multi-view videos for sufficient geometric constraints, making capture expensive and limiting sparse-camera deployment. Reducing input views lowers acquisition cost but weakens geometry supervision, often causing missing structures and floating Gaussians. Depth priors provide geometric cues, yet no single source offers both dense coverage and reliable geometry. Monocular depth provides dense structure but is scale-ambiguous and locally biased, whereas multi-view geometric depth provides incomplete anchors consistent with the reconstruction coordinate system. To exploit their complementarity, we propose D$^2$-4DGS, a sparse-camera dynamic 4D Gaussian Splatting framework guided by dual-source depth priors. We align monocular estimates with valid multi-view geometric depths and verify their consistency to identify reliable geometric anchors. These verified anchors support consistency-aware pruning and depth supervision, while verified geometric depths and aligned mono-only estimates provide candidate geometry for densification in under-reconstructed regions. Finally, RGB-D joint optimization improves appearance fidelity and geometric consistency under sparse-view supervision. Across all nine dataset--view settings, D$^2$-4DGS achieves the highest PSNR, improving by 1.33 dB on average over the best competing method in each setting.