Spatial Filter
Radius
Defines the width of the spatial filter. Higher values lead to stronger noise reduction at the cost of small details and performance.
Performance impact: moderate, don't increase above the necessary level
Spatial Occlusion ReSTIR Validation plays a very important role in preserving details during spatial filtering. Without it, over-blurring of small features (especially indirect shadows) becomes much more noticeable.
The comparison below demonstrates the same Radius values, with Spatial Occlusion ReSTIR Validation turned off:
Adaptivity
Defines the adaptive behavior of the spatial filter. Higher values allow the filter radius to shrink more in corners and crevices (where small-scale indirect shadowing and light bounce are expected), better preserving details at the cost of filtering strength.
Performance impact: small
Adaptivity, while efficient at preserving small-scale details, cannot replace Spatial Occlusion ReSTIR Validation. The best results are achieved when the two are used together:
Recurrent Blur
Enables an algorithm that combines spatial and temporal denoising stages in a way that achieves stronger spatial blurring without a large performance impact. This comes at the cost of losing some details and slightly reducing temporal responsiveness.
Performance impact: small
Firefly Suppression
Enables a filter that removes bright outliers ("fireflies") from the tracing result before the denoiser runs. This reduces boiling artifacts and improves temporal stability, at the cost of some image darkening.
Performance impact: small
Last updated