分裂辐射级联：通过稀疏辐射探针实现实时全局光照
Split Radiance Cascades: Real-Time Global Illumination via Sparse Radiance Probes
浏览论文内容
中文总结 AI 辅助
研究针对辐射探针方法缺乏细节的问题，采用稀疏哈希表存储世界空间探针，引入光线分裂计算辐射区间，实现分裂辐射级联算法，能在多种场景下为实时3D漫反射全局光照提供高质量间接光照。
中文摘要 AI 辅助
辐射探针方法是用于实时图形中近似漫反射全局光照的常用且经过充分测试的方法，但由于探针间距大，通常缺乏细节。辐射级联（RC）通过提高空间分辨率并降低来自较近物体的光和遮挡的角度分辨率来解决此问题，从而能在所有尺度上提供无噪声或混叠的细节。然而，由于存储高细节体辐射信息成本过高，RC的主流实现要么在二维或屏幕空间中运行。在这项工作中，我们采用稀疏哈希表存储世界空间探针，使辐射级联适用于精确的实时3D漫反射全局光照。我们引入光线分裂，一种通过从可见表面追踪光线并根据其命中距离计算它们对级联的贡献来计算RC中使用的辐射区间的方法。我们在各种场景中评估了我们的算法——分裂辐射级联，并证明它能在单帧和时间累积环境中都提供高质量的间接光照。
英文摘要
Radiance probe methods are a popular and well-tested approach for approximating diffuse global illumination for real-time graphics, but they commonly suffer from a lack of detail due to the large spacing between probes. Radiance Cascade (RC) fixes this by increasing spatial resolution and reducing angular resolution for light and occlusion from closer objects, which allows it to provide details at all scales without noise or aliasing. However, leading implementations of RC either run in 2D or screenspace, due to the prohibitive costs of storing high-detail volumetric radiance information. In this work, we adapt Radiance Cascades for accurate real-time 3D diffuse global illumination using a sparse hashmap to store world-space probes. We introduce ray splitting, a method for calculating radiance intervals used in RC by tracing rays from visible surfaces and calculating their contribution to cascades based on their hit distance. We evaluate our algorithm, Split Radiance Cascades, on a variety of scenes, and demonstrate that it can provide high-quality indirect illumination in both single-frame and temporally accumulated contexts.
