面向高保真 3D 高斯溅射的深度-边缘感知正则化

3D 高斯溅射 (3DGS) 表示在 efficient and high-fidelity novel view synthesis 领域的 significant advancement。Despite recent progress，achieving accurate geometric reconstruction under sparse-view conditions 仍是 a fundamental challenge。Existing methods often rely on non-local depth regularization，which fails to capture fine-grained structures and is highly sensitive to depth estimation noise。Furthermore，traditional smoothing methods neglect semantic boundaries and indiscriminately degrade essential edges and textures，consequently limiting overall quality of reconstruction。在本 work 中，我们提出 DET-GS，一个 unified depth and edge-aware regularization framework for 3D 高斯溅射。DET-GS 引入 hierarchical geometric depth supervision framework adaptive enforce multi-level geometric consistency，显著 enhance structural fidelity and robustness against depth estimation noise。To preserve scene boundaries，我们 design edge-aware depth regularization guided by semantic masks derived from Canny edge detection。Furthermore，我们 introduce RGB-guided edge-preserving Total Variation loss selectively smooth homogeneous regions while rigorously retaining high-frequency details and textures。Extensive experiments demonstrate that DET-GS achieves substantial improvements in both geometric accuracy and visual fidelity，outperforming state-of-the-art (SOTA) methods on sparse-view novel view synthesis benchmarks。