Path guiding for path-traced rendering

Abstract

A computer-implemented method for generating a mask for a light source in a virtual scene includes determining a bounding box for the scene based on a frustum of a virtual camera and generating a path-traced image of the scene within the bounding box. Light paths emitted by the camera and exiting at the light source are stored, and objects poorly sampled by the light source are removed from the scene. An initial mask for the light source is generated from the density of light paths exiting at that position on the light source. The initial mask is refined by averaging in the light path density at each point on the light source for subsequent images.

What is claimed is: 
     
         1 . A computer-implemented method for generating a light guide path for use in rendering virtual scenes, the computer-implemented method comprising:
 under control of one or more computer systems configured with executable instructions, perform steps of:
 placing within a scene a virtual camera having a virtual camera position; 
 generating a masked light source data structure for a light source of the scene by assigning an emission probability to positions on the light source based on a density of light paths emitted by the virtual camera exiting at that position on the light source or of light paths emitted from that position on the light source and exiting at the virtual camera position; 
 generating a path-traced image by using the masked light source data structure to generate a plurality of light paths and storing a light path subset comprising the light paths of the plurality of light paths that exit at the virtual camera; 
 identifying a light paths subset of the plurality of light paths that are sampled less than a pre-determined threshold; 
 for at least some light paths of the light paths subset:
 computing a distribution for at least one vertex of a first plurality of vertices of the light path; 
 using the distributions, creating a second plurality of vertices of the light path by sampling a position for at least one other vertex of the first plurality of vertices; 
 constructing a guided light path from the second plurality of vertices; and 
 modifying the guided light path by iteratively sampling the positions of the second plurality of vertices until the guided light path is visible to the virtual camera; and 
 
 forming a scene image storable in computer memory by adding the modified guided light paths to the path-traced image. 
   
     
     
         2 . The computer-implemented method of  claim 1 , wherein the distribution of each vertex of the second plurality of vertices comprises a Gaussian distribution. 
     
     
         3 . The computer-implemented method of  claim 1 , wherein the distribution of each vertex of the second plurality of vertices comprises an anisotropic reconstruction kernel. 
     
     
         4 . The computer-implemented method of  claim 1 , further comprising: if a termination/splitting factor is greater than 1, splitting the guided light path, at a current vertex of the second plurality of vertices, into at least two guide paths. 
     
     
         5 . The computer-implemented method of  claim 4 , wherein splitting the guided light path comprises resampling at least one vertex of the second plurality of vertices, after the current vertex, for each distribution. 
     
     
         6 . The computer-implemented method of  claim 1 , further comprising: if a termination/splitting factor is less than 1, terminating the guided light path at a current vertex of the second plurality of vertices. 
     
     
         7 . The computer-implemented method of  claim 6 , wherein terminating the guided light path comprises deleting at least some vertices of the guided light path. 
     
     
         8 . The computer-implemented method of  claim 1 , wherein the guided light paths comprise diffuse reflection, diffuse refraction, or caustics. 
     
     
         9 . The computer-implemented method of  claim 8 , further comprising denoising the scene image. 
     
     
         10 . The computer-implemented method of  claim 9 , wherein the denoising comprises density-based outlier rejection. 
     
     
         11 . The computer-implemented method of  claim 1 , wherein identifying the light paths subset of the plurality of light paths that are sampled less than a pre-determined threshold involves the use of quad trees. 
     
     
         12 . A computer system for modeling surfaces of virtual objects, the system comprising:
 at least one processor; and   a computer-readable medium storing instructions, which when executed by the at least one processor, causes the system to carry out the method of  claim 1 .   
     
     
         13 . A non-transitory computer-readable storage medium storing instructions, which when executed by at least one processor of a computer system, causes the computer system to carry out the method of  claim 1 .