Metropolis Light Transport for Participating Media Mark Pauly

  • Slides: 17
Download presentation
Metropolis Light Transport for Participating Media Mark Pauly ETH Zürich Thomas Kollig Alexander Keller

Metropolis Light Transport for Participating Media Mark Pauly ETH Zürich Thomas Kollig Alexander Keller University of Kaiserslautern EGWR 2000

Overview l Light Transport for Participating Media l Path Integral Formulation l Sampling l

Overview l Light Transport for Participating Media l Path Integral Formulation l Sampling l Rendering with Metropolis Light Transport l Results l Conclusions EGWR 2000

Related Work MC Methods FE Methods l Zonal Methods ‘ 87 l Rushmeier, Torrance

Related Work MC Methods FE Methods l Zonal Methods ‘ 87 l Rushmeier, Torrance l Hierarchical Radiosity ‘ 93 Pattanaik, Mudur l Spherical Harmonics ‘ 84 l Discrete Ordinates ‘ 94 Languenou, Bouatouch, Chelle Photon Map ‘ 98 Jensen, Christensen Kajiya, von Herzen l Bidirectional Path Tracing ‘ 96 Lafortune, Willems Bhate l Light Tracing ‘ 93 l Metropolis Light Transport ‘ 97 Veach, Guibas EGWR 2000

Light Transport l Global Balance Equation Streaming Emission Absorption In-scattering Out-scattering EGWR 2000

Light Transport l Global Balance Equation Streaming Emission Absorption In-scattering Out-scattering EGWR 2000

Path Integral Formulation l Measurement Equation l Path Integral EGWR 2000

Path Integral Formulation l Measurement Equation l Path Integral EGWR 2000

l Path Characteristic 1 0 medium 1 sensor object 1 light source EGWR 2000

l Path Characteristic 1 0 medium 1 sensor object 1 light source EGWR 2000

l Path Space Measure EGWR 2000

l Path Space Measure EGWR 2000

l Measurement Contribution Function l Path Integral EGWR 2000

l Measurement Contribution Function l Path Integral EGWR 2000

Sampling l Line Integral Computation: Ray Marching Equidistant Sampling Stratified Sampling efficient aliasing anti-aliasing

Sampling l Line Integral Computation: Ray Marching Equidistant Sampling Stratified Sampling efficient aliasing anti-aliasing inefficient Random Offset Sampling EGWR 2000

Metropolis Light Transport l Generate a random walk path space through l For each

Metropolis Light Transport l Generate a random walk path space through l For each path deposit a constant amount of energy at the corresponding pixel l Obtain desired image by distributing paths according to image contribution Metropolis sampling EGWR 2000

l Metropolis Sampling l Propose a mutation of current path l Compute acceptance probability

l Metropolis Sampling l Propose a mutation of current path l Compute acceptance probability l Choose as new sample if Samples are correlated we can exploit coherence EGWR 2000

l Mutation Strategies l Bidirectional Mutations l large changes to the current path ensures

l Mutation Strategies l Bidirectional Mutations l large changes to the current path ensures ergodicity Perturbations l l l high acceptance probability changes to image location low cost Scattering Perturbations Sensor Perturbations Propagation Perturbations Caustic Perturbations EGWR 2000

l Propagation Perturbation medium image plane eye light source EGWR 2000

l Propagation Perturbation medium image plane eye light source EGWR 2000

Results EGWR 2000

Results EGWR 2000

Results EGWR 2000

Results EGWR 2000

Results EGWR 2000

Results EGWR 2000

Conclusions l Participating media are fully integrated l l inhomogeneous media multiple, anisotropic scattering

Conclusions l Participating media are fully integrated l l inhomogeneous media multiple, anisotropic scattering volume caustics color bleeding General geometry and reflection models l Robust l l l Complex Scenes Difficult Lighting Situations EGWR 2000