We extend a recent direct illumination sampling technique, spatiotemporal reservoir resampling (ReSTIR), to multi-dimensional path space for volumetric media. By fully evaluating just a single path sample per pixel, our volumetric path tracer shows unprecedented convergence. To achieve this, we properly estimate the chosen sample's probability via approximate perfect importance sampling with spatiotemporal resampling. With this reformulation, we achieve low-noise, interactive volumetric path tracing with arbitrary dynamic lighting, including volumetric emission, and maintain interactive performance even on high-resolution volumes.
Baseline: a fast implementation of decomposition tracking [Kutz et al. 2017] (to sample free flight distances) and residual ratio tracking [Novák et al. 2014] (for estimating transmittance in next event estimation).
High-quality version of the paper video (822 MB).