Area ReSTIR: Resampling for Real-Time Defocus and Antialiasing

Abstract

Recent advancements in spatiotemporal reservoir resampling (ReSTIR) leverage sample reuse from neighbors to efficiently evaluate the path integral. Like rasterization, ReSTIR methods implicitly assume a pinhole camera and evaluate the light arriving at a pixel through a single predetermined subpixel location at a time (e.g., the pixel center). This prevents efficient path reuse in and near pixels with high-frequency details.

We introduce Area ReSTIR, extending ReSTIR reservoirs to also integrate each pixel’s 4D ray space, including 2D areas on the film and lens. We design novel subpixel-tracking temporal reuse and shift mappings that maximize resampling quality in such regions. This robustifies ReSTIR against high-frequency content, letting us importance sample subpixel and lens coordinates and efficiently render antialiasing and depth of field.

Path Tracing vs. ReSTIR PT [Lin et al. 2022] vs. Area ReSTIR (Ours)

Prior ReSTIR methods [Bitterli et al. 2020] reuse paths between fixed subpixel locations. Randomizing paths’ subpixel locations avoids aliasing, but changes their primary hits. In pixels with high-frequency content, the changing normals and occlusions often make the new primary hits incompatible with later vertices (from reused, shifted paths), preventing effective reuse. Here the Franck model, with complex fur, is lit by the Blue Lagoon HDR environment. (a) Path tracing has high variance at low sample rates. On smoother surfaces, (b) ReSTIR PT [Lin et al. 2022] reuses samples for large quality improvements while shading only one path per pixel. But reusing samples in a higher-dimensional ray space, including subpixel and lens coordinates, allows our (c) Area ReSTIR to reuse even more paths (despite bokeh and high-frequency content) also while shading just one sample per pixel. This approaches the quality of (d) converged path tracing. Note: (a), (b), and (c) all use the same number of independent paths each pixel.

Video

Video Summary

BibTeX

@article{Zhang*2024,
   author       = {Song Zhang* and Daqi Lin* and Markus Kettunen and Cem Yuksel and Chris Wyman},
   title        = {Area ReSTIR: Resampling for Real-Time Defocus and Antialiasing},
   journal      = {ACM Transactions on Graphics (Proceedings of SIGGRAPH 2024)},
   year         = {2024},
   month        = {07},
   volume       = {43},
   number       = {4},
   pages        = {98:1--98:13},
   articleno    = {98},
   numpages     = {13},
   location     = {Denver, CO, USA},
   url          = {https://doi.org/10.1145/3658210},
   doi          = {10.1145/3658210},
   issn         = {0730-0301},
   publisher    = {ACM Press},
   address      = {New York, NY, USA},
   note         = {(*Joint First Authors)},
}

• ← Previous Publication