NVIDIA-RTX/NRD

NVIDIA Real-time Denoising (NRD) library

What it solves

NRD provides a high-quality, real-time denoising solution for noisy signals in graphics rendering, specifically targeting 1 path/pixel path tracing. It addresses the noise inherent in real-time ray tracing by resolving it on opaque surfaces using G-buffer guides (normals, roughness, viewZ, and motion vectors), supporting both static and dynamic lighting.

How it works

NRD is an API-agnostic, spatio-temporal library that integrates into rendering engines via D3D12, Vulkan, or D3D11. It uses per-pixel G-buffer guides to resolve noise. The library includes three specialized denoisers:

  • REBLUR: A recurrent blur-based denoiser for diffuse and specular radiance, as well as occlusion.
  • RELAX: An A-trous based denoiser designed for RTX Direct Illumination (RTXDI), handling diffuse and specular radiance.
  • SIGMA: A per-light shadow-only denoiser for infinite or local light sources, including shadows with translucency.

It offers a "SH" (Spherical Gaussian) mode for higher quality results comparable to DLSS-RR, serving as a non-AI alternative.

Who it’s for

  • Game developers creating AAA titles.
  • Developers of professional visualization (ProVis) applications (e.g., Autodesk Aurora, Enscape, Lumion).
  • Graphics engineers integrating real-time denoising into D3D12, Vulkan, or D3D11 engines.

Highlights

  • API Agnostic: Works across D3D11, D3D12, and Vulkan.
  • High Performance: Optimized for RTX hardware (e.g., REBLUR diffuse/specular takes ~2.55ms on an RTX 4080 at 1440p).
  • Versatile Denoisers: Includes specialized tools for radiance (REBLUR, RELAX) and shadows (SIGMA).
  • Non-AI Alternative: Provides high-quality results without relying on AI-based upscaling or denoising.

Related

  • Project
    RenderKit/oidn

    A high-performance library of deep learning-based denoising filters that reduces rendering times for ray tracing applications by removing Monte Carlo noise.

  • Project
    nvpro-samples/vk_gltf_renderer

    A Vulkan RTX path tracer and glTF 2.0 PBR material reference featuring a built-in scene editor and AI-powered denoising and generation tools.

  • Project
    nachifur/RDDM

    A residual denoising diffusion model framework for high-quality image restoration and generation, supporting tasks like deraining and deblurring.

  • Project
    nerfstudio-project/gsplat

    A CUDA-accelerated library for the rasterization of 3D Gaussians, providing a faster and more memory-efficient way to render radiance fields compared to the official implementation.

  • Project
    nvidia-isaac/cuVSLAM

    A CUDA-accelerated library for visual odometry and SLAM that provides real-time, high-accuracy spatial tracking across various camera and IMU configurations without requiring parameter tuning.