Abstract
Recent extensions to spatiotemporal path reuse, or ReSTIR, improve rendering efficiency in the presence of high-frequency content by augmenting path reservoirs to represent contributions over full pixel footprints. Still, if historical paths fail to contribute to future frames, these benefits disappear. Prior ReSTIR work backprojects to the prior frame to identify paths for reuse. Backprojection can fail to find relevant paths for many reasons, including moving cameras or subpixel geometry with differing motion. We introduce reservoir splatting to reduce these failures. Splatting forward-projects the primary hits of prior-frame paths. Unlike backprojection, forward-projected path samples fall into the current-frame pixel relevant to their exact primary hits, making successful reuse more likely. This also enables motion blur for ReSTIR, by splatting at multiple time steps, and supports depth of field without the specialized shift maps needed previously.Beyond enabling motion blur, splatting improves resampling quality over Zhang et al.’s [2024] Area ReSTIR at up to 10 % lower cost. To improve robustness, we show how to MIS splatted and backprojected samples to help every current-frame pixel get at least one historical path proposed for reuse.
Prior ReSTIR methods [Zhang et al. 2024] degrade during camera motion, as sequential frames rarely shade identical primary hits, making perfect reuse tricky. This worsens near fine details, like foliage and fur, where sequential frames may not hit the same surface. By forward splatting hits from last frame, we guarantee they are reevaluated. Adding time to samples also enables resampling for motion blur while shading only one sample per pixel. Here we show two scenes under camera motion (Sheep In Forest and Subway) with stock Area ReSTIR [Zhang et al. 2024] and our new splatting-based ReSTIR. Insets also show an offline reference and a naïve motion blur baseline using Zhang et al.’s [2024] backprojection.
Video
BibTeX
@inproceedings{liu2025splatting,
title = {Reservoir Splatting for Temporal Path Resampling and Motion Blur},
author = {Liu, Jeffrey and Lin, Daqi and Kettunen, Markus and Wyman, Chris and Ramamoorthi, Ravi},
month = {August},
journal = {SIGGRAPH (Conference Track)},
year = {2025},
doi = {10.1145/3721238.3730646},
}