面向高保真高斯增强面元渲染的深度剥离
Depth Peeling for High-Fidelity Gaussian-Enhanced Surfel Rendering
用 Depth Peeling 为 2D 不透明面元做逐像素精确排序，从而在无排序的高斯泼溅框架下正确调制透射率，消除走样与 popping；这套「面元+边界半透明+周围高斯补细节」的混合表示可直接迁移到任何需要正确混合顺序的实时神经渲染管线。
原文依据 We employ Depth Peeling to achieve accurate per-pixel ordering for surfel rendering, which enables sort-free Gaussian splatting with correct transmittance modulation
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论NeRF 与 3D 高斯泼溅（3DGS）显著推进了新视角合成，但它们需要对体采样点或图元进行排序才能正确混合颜色。近期的 Gaussian-Enhanced Surfels（GES）实现了高性能的无排序渲染，却受走样伪影与重建欠佳的困扰。为克服这些局限，我们提出 DP-GES，一种新的表示：用带半透明边界的 2D 不透明面元表示粗尺度几何与外观，并用面元周围的 3D 高斯补充细尺度细节。我们采用深度剥离为面元渲染实现逐像素的精确排序，从而使无排序的高斯泼溅能够进行正确的透射率调制，有效消除走样与 popping 伪影，并支持完全可微的联合优化。大量实验表明，本方法取得了更优的新视角合成质量，并在多种场景下优于最优技术。
英文原文摘要 1075 字符
Novel view synthesis has been significantly advanced by NeRFs and 3D Gaussian Splatting (3DGS), which require ordering volumetric samples or primitives for correct color blending. While the recent Gaussian-Enhanced Surfels (GES) enable high-performance, sort-free rendering, they suffer from aliasing artifacts and suboptimal reconstruction. To address these limitations, we propose DP-GES, a novel representation that combines 2D opaque surfels with semi-transparent boundaries to represent coarse-scale geometry and appearance, and 3D Gaussians surrounding the surfels to supplement fine-scale details. We employ Depth Peeling to achieve accurate per-pixel ordering for surfel rendering, which enables sort-free Gaussian splatting with correct transmittance modulation, effectively eliminating aliasing and popping artifacts while facilitating a fully differentiable joint optimization. Extensive experiments demonstrate that our method achieves superior novel view synthesis quality and compares favorably against state-of-the-art techniques across a wide range of scenes.