Generalized resampled importance sampling

As scenes become ever more complex and real-time applications embrace ray tracing, path sampling algorithms that maximize quality at low sample counts vital. Recent resampling building on Talbot et al.'s [2005] resampled importance (RIS) reuse paths spatiotemporally to render surprisingly light transport with a few samples per pixel. These reservoir-based spatiotemporal resamplers (ReSTIR) their underlying RIS theory make various assumptions, including independence. But introduces correlation , so ReSTIR-style iterative loses most convergence guarantees theoretically provides. We introduce generalized (GRIS) extend the theory, allowing correlated samples, unknown PDFs taken from varied domains. This solidifies theoretical foundation, us derive variance bounds conditions in ReSTIR-based samplers. It also guides practical algorithm design enables advanced between pixels via shift mappings. show path-traced resampler (ReSTIR PT) running interactively scenes, capturing many-bounce diffuse specular lighting while shading just one With our new we can modify guarantee for offline renderers.