Occlusion-Point Reuse for Ray-Traced Ambient Occlusion and Shadow

A new occlusion-point reuse framework targets two of the most common visibility effects in real-time rendering: ambient occlusion and area-light shadows. Instead of reusing finished shading values or light samples, it reuses traced samples in the domain of the first-hit occlusion point, then combines neighboring occluder samples with multiple importance sampling.

That shift matters because hardware ray tracing still leaves games with very few rays per pixel. In that regime, raw AO and shadow estimates are noisy, and the usual fixes often trade quality for stability. Reframing the problem around occluder-domain integrals gives the method a more direct handle on visibility, especially for soft shadows where local occluder structure can be a better reuse signal than visibility consistency alone.

The work includes both unbiased estimators, which converge to the transformed integrals, and biased versions intended for practical real-time use. The biased shadow variant assumes local first-hit occluder consistency, and the authors argue that this better matches local visibility geometry than reusing light samples. For teams shipping ray-traced lighting, that’s a useful reminder that reuse strategy can matter as much as denoising.

In experiments, the method produced higher-quality AO and shadows than non-reuse ray-traced baselines, and better shadow quality than light-sample reuse at comparable cost. The exact engine integration path hasn’t been disclosed, but the technique is directly relevant to graphics programmers working on ray-traced visibility, temporal reuse, and...