Using previously rendered scene frames to reduce pixel noise

Patent metadata

Field	Value
Publication number	US-2021264571-A1
Application number	US-202117241782-A
Country	US
Kind code	A1
Filing date	Apr 27, 2021
Priority date	Aug 14, 2018
Publication date	Aug 26, 2021
Grant date	—

Abstract

This disclosure presents a method and computer program product to denoise a ray traced scene. An apparatus for processing a ray traced scene is also disclosed. In one example, the method includes: (1) generating filtered scene data by filtering modified scene data from original scene data utilizing a spatial filter, and (2) providing a denoised ray traced scene by adjusting the filtered scene data utilizing a temporal filter. The modified and adjusted scene data can be sent to a rendering processor or system to complete rendering to generate a final scene.

First claim

What is claimed is:

1 . A method of denoising a ray traced scene, comprising:
generating filtered scene data by filtering modified scene data from original scene data of a ray traced scene utilizing a spatial filter; and providing a denoised ray traced scene by adjusting the filtered scene data utilizing a temporal filter.

2 . The method as recited in claim 1 , further comprising generating the modified scene data by modifying the original scene data utilizing an exponential average of temporal reprojections.

3 . The method as recited in claim 2 , wherein utilizing the exponential average includes clamping the temporal reprojections using a color spatial mean and a color standard deviation associated with a pixel window of the original scene data.

4 . The method as recited in claim 1 , wherein utilizing the spatial filter includes reducing a spatial filter radius based on a number of a temporal reprojections within the spatial filter radius.

5 . The method as recited in claim 4 , further comprising determining the number by dividing the spatial filter radius by a square root of the number of temporal reprojections within the spatial filter radius, wherein the square root of the number of temporal reprojections is clamped to a maximum value.

6 . The method as recited in claim 1 , wherein adjusting the filtered scene data further comprises determining a temporal filter radius for the temporal filter that is larger than a spatial filter radius utilized with the spatial filter and clamping the temporal filter radius utilizing a color spatial mean and a color standard deviation associated with the original scene data.

7 . The method as recited in claim 1 , wherein filtering the modified scene data further comprises scaling a spatial filter radius utilized with the spatial filter using a global illumination history count.

8 . The method as recited in claim 1 , further comprising preparing the original scene data before the modifying by applying a spatial denoiser to the scene data utilizing a denoise radius.

9 . The method as recited in claim 8 , wherein the denoise radius is based on a set of caustic photons from a light source that are positioned on a surface of an object in the ray traced scene and a logarithmic cost of the set of caustic photons that intersect pixels of a pixel window of the original scene data.

10 . The method as recited in claim 1 , further comprising receiving the original scene data, wherein the original scene data corresponds to a scene from a game application.

11 . An apparatus for processing a ray traced scene, comprising:
a processor that generates modified scene data from original scene data utilizing an exponential average derived from temporal reprojections, filters the modified scene data utilizing a spatial filter, and adjusts the filtered scene data utilizing a temporal filter.

12 . The apparatus as recited in claim 11 , further comprising a processor that computes a logarithmic cost utilizing an iterative count of caustic photons, and applies the spatial filter to the modified scene data using a spatial filter radius modified by the logarithmic cost.

13 . The apparatus as recited in claim 11 , further comprising a processor that renders a scene employing the modified scene data.

14 . The apparatus as recited in claim 11 , wherein the processor utilizes a spatial filter radius with the spatial filter and reduces the spatial filter radius based on a number of a temporal reprojections within the spatial filter radius.

15 . The apparatus as recited in claim 11 , wherein the processor adjusts the filtered scene data utilizing a temporal filter radius for the temporal filter that is larger than a spatial filter radius utilized with the spatial filter.

16 . The apparatus as recited in claim 11 , wherein the processor prepares the original scene data by applying a spatial denoiser to the scene data utilizing a denoise radius.

17 . The apparatus as recited in claim 11 , wherein the processor is a graphics processing unit (GPU).

18 . A computer program product having a series of operating instructions stored on a non-transitory computer-readable medium that directs a data processing apparatus when executed thereby to perform operations to denoise a ray traced scene, comprising:
generating modified scene data by modifying scene data from an application; generating filtered scene data by filtering the modified scene data utilizing a spatial filter; and providing a denoised ray traced scene by adjusting the filtered scene data utilizing a temporal filter.

19 . The computer program product as recited in claim 18 , wherein adjusting the filtered scene data utilizes a temporal filter radius for the temporal filter that is larger than a spatial filter radius utilized with the spatial filter.

20 . The computer program product as recited in claim 18 , further comprising receiving the scene from the application, wherein the application is a game application and the operations are performed on one or more graphics processing units (GPU).

Assignees

• Nvidia Corp

Inventors

• Liu Shiqiu
• Pantaleoni Jacopo

Classifications

• G06T15/50

  Lighting effects · CPC title

• G06T15/06 Primary

  Ray-tracing · CPC title

• G06T5/10

  using non-spatial domain filtering · CPC title

• G06T5/70 Primary

  Denoising; Smoothing · CPC title

• G06T5/40

  using histogram techniques · CPC title