用可微分的 Voronoi 射线追踪实现实时渲染，支持多种镜头特效。
Differentiable Voronoi Ray Tracing Beyond Rasterization Speeds
- 设计新型场景表示与 GPU 执行协同优化，降低射线追踪开销。
- 在 Mip-NeRF 360 上达 623 FPS，速度超前代方法 3.2 倍。
- 无需专用光栅化，自然支持鱼眼、滚动快门等复杂效果。
实时新视角合成主要依赖光栅化的显式基元。这类基于投影的流程虽有高吞吐量，但难以处理非针孔效应（如畸变、滚动快门、景深）。基于射线的渲染能自然表达这些效应，但通常被认为不适合实时竞争。我们分析了可微分 Voronoi 射线追踪的吞吐量影响因素，识别出遍历长度、每单元计算量和内存局部性为关键决定因素。据此提出 VoroTracing，通过联合设计场景表示、优化策略与 GPU 执行来降低上述成本。紧凑的八面体外观纹理减少内存流量，表面集中透明度促进早期终止。固定预算表示无需剪枝或加密，且使用针对相干遍历优化的 GPU 实现。在 Mip-NeRF 360 上，VoroTracing 在 RTX 5090 上实现 623 FPS，吞吐量为最快先前射线方法的 3.2 倍，比 3D 高斯溅射高 2.8 倍，同时保持竞争力的重建质量。渲染器通过射线生成与采样支持鱼眼、滚动快门、运动模糊和景深，无需专用光栅化。结果表明，射线渲染可实现实时吞吐量，兼具灵活性。代码已公开。
原文摘要 · Abstract (English)
Real-time novel view synthesis is dominated by rasterized explicit primitives. These projection-based pipelines provide high throughput but require specialized extensions for non-pinhole effects such as distortion, rolling shutter, and depth of field. Ray-based rendering expresses these effects naturally but is generally assumed too slow for competitive real-time rendering. We analyze the factors governing throughput in differentiable Voronoi ray tracing and identify traversal length, per-cell work, and memory locality as principal determinants. Guided by this, we introduce VoroTracing, which co-designs the scene representation, optimization, and GPU execution to reduce these costs. Compact octahedral appearance textures reduce memory traffic, while surface-concentrated opacity promotes early termination. The fixed-budget representation is optimized without pruning or densification and rendered with a GPU implementation designed for coherent traversal. On Mip-NeRF 360, VoroTracing renders at 623 FPS on an RTX 5090, providing $3.2\times$ the throughput of the fastest prior ray-based method and $2.8\times$ that of 3D Gaussian Splatting, while maintaining competitive reconstruction quality. Our renderer supports fisheye, rolling-shutter, motion-blur, and depth-of-field effects through ray generation and sampling, requiring no specialized rasterization. These results show that real-time throughput can be achieved with the flexibility of ray-based rendering. We release our source code, see https://research.zenseact.com/publications/vorotracing
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。