Overview
ReSTIR keeps weighted sample reservoirs per pixel and shares them across time or space so difficult lighting can be sampled more efficiently.
SpecializedSource checked
ReSTIR Sampling
Reuses light samples with reservoir sampling to improve direct lighting estimates.
Quick Facts
Compact metadata for implementation and discovery.
- Category
- Rendering
- Difficulty
- Advanced
- Family
- Light Transport Sampling
- Tier
- Specialized
- Status
- Verified
- Review
- Source Checked
- Sources
- 1
- Formulas
- 1
- Relationships
- 2
- Code example
- Ready
- Visualization
- Ready
- Aliases
- Reservoir Spatio-Temporal Importance Resampling
Core Concepts
reservoir samplingimportance weightspatiotemporal reusedirect lighting
Quick facts
Complexity
Runtime and memory behavior at a glance.
Best
O(samples)
Average
O(pixels * candidates)
Worst
O(pixels * candidates)
Space
O(pixels)
When to Use
- Use when real-time graphics needs reservoir sampling.
- Use when the input matches the assumptions behind light Transport Sampling.
When Not to Use
- Avoid when the effect costs more than the visual improvement for the target hardware.
- Avoid when a simpler algorithm gives the required result with less setup.
What It Does
- Reuses light samples with reservoir sampling to improve direct lighting estimates
- Uses reservoir sampling and importance weight to guide the work.
- Produces rendered image.
Use Cases
- Real-time graphics
- Offline rendering
- Lighting and visibility effects
Inputs
- Scene geometry
- Materials
- Camera and lighting state
Outputs
- Rendered image
- Visibility or shading buffers
Key Formulas (1)
Reservoir weight
W = sumweights / selectedweight
W = sum_weights / selected_weightTracks how a selected sample represents the full candidate set.
Pseudocode
- 1.Prepare the inputs and state needed for ReSTIR Sampling.
- 2.Apply the reservoir sampling step while the stopping condition has not been met.
- 3.Update intermediate state and discard invalid candidates.
- 4.Return rendered image.
Canonical Example
- Input
- ReSTIR Sampling input with a small representative case
- Output
- Rendered image
- Explanation
- ReSTIR Sampling applies reservoir sampling to produce rendered image.
Implementation Notes
- Keep the data space for positions, normals, and samples explicit in the implementation.
- Keep the reservoir sampling step separate enough to test independently.
Common Pitfalls
- Mixing coordinate spaces or sampling domains leads to artifacts that are hard to debug.
- Forgetting to check edge cases can make ReSTIR Sampling look correct on only ideal inputs.
Interactive Mechanism Walkthrough
Spatial Primitive & Sweep-Line CanvasShow rays, samples, buffers, or shading terms that contribute to the final pixel.
Step 1 of 4: Prepare the inputs and state needed for ReSTIR Sampling.
rendering-mechanismNotes
ReSTIR depends on reservoir weights and reuse mappings; biased variants reduce noise but can lose energy.
Sources (1)
What each reference supports for this entry.
ReSTIR paper
PaperNVIDIA research page for spatiotemporal reservoir resampling for dynamic direct lighting.
Open sourceExplore Next
Good next jumps from the relationship graph, ranked for useful discovery.
Relationships
View full relationship pageImproves onImportance Sampling BRDF
ReSTIR reuses reservoirs to improve difficult light sampling scenarios.
Often paired withPath Tracing
ReSTIR is commonly applied in real-time path tracing pipelines.
Related Algorithms
Tags
GraphicsSamplingLightingGlobal Illumination