Fast Global Illumination Including Specular Effects Xavier Granier 1 George Drettakis 1 Bruce J. Walter 2 1 iMAGIS -GRAVIR/IMAG-INRIA iMAGIS is a joint project of CNRS/INRIA/UJF/INPG 2 Cornell University iMAGIS EGWR /06/2000 Motivation Realistic Illumination All light paths Time-Quality Tradeoff Interactive Visualisation Quality Control iMAGIS EGWR /06/2000 Talk overview Previous work New Integrated Algorithm Results Conclusion iMAGIS EGWR /06/2000 Previous Work Deterministic methods Radiosity [Goral84,Cohen88,etc] Hierarchy and Clustering [Hanrahan91, Smits94, Sillion95, etc] Non diffuse [ Immel86, Sillion89, Sillion91, etc] Probabilistic Photon Map [Jensen96,etc] Density Estimation [Walter97,etc] iMAGIS EGWR /06/2000 Previous Work Multi-pass Two-pass [Wallace97,Sillion89,etc] Integrated [Chen91,etc] Interactive viewing Render-Cache [Walter99] Directional Storage [Stamminger99,etc] iMAGIS EGWR /06/2000 Overview DD transfer Hierarchical Radiosity with Clustering (HRC) DS + D transfer Particle tracing during HRC gather D = Diffuse and S = Non Diffuse Images have specular path to eye added by Ray-Tracing iMAGIS EGWR /06/2000 Algorithm Overview Construct hierarchy Hierarchy elements: clusters and surfaces For each iteration Refine create links at correct level Gather - Energy transfer particle emission restricted by links Push-pull particle placement iMAGIS EGWR /06/2000 Refinement Link placement Choose appropriate hierarchy level for transfer Refinement test: Energy > Visibility classification and computation Shafts and blocker lists for classification/optimisation Unoccluded form factor computation iMAGIS EGWR /06/2000 I RS = Radiosity x Form Factor x Visibility Energy transfer through a link Diffuse-Diffuse transfer I RS I R = I R + I RS iMAGIS EGWR /06/2000 Energy transfer through links Diffuse-Specular transfer Diffuse-Specular transfer Probabilistic emission of particles Reflection on receiver Propagation and impact storage Links guide particles Links encode light flow Restrict number of particles iMAGIS EGWR /06/2000 Particle Emission Number of particles Flux S to R / Constant energy Uniform sampling Inverse of ( Measure(R) x Measure(S) ) Particle power Flux from s to r corrected by number of particles and probability of sample choice iMAGIS EGWR /06/2000 Push-Pull Push: Hierarchy descent Particle placement Integrate particle power into irradiance Radiosity computation on leaves Pull: Radiosity averaging iMAGIS EGWR /06/2000 Particle Placement Detect high variation and concentration Quantity Average position and "Spread Factor" Push particle if: High concentration and high energy iMAGIS EGWR /06/2000 Interactive Visualisation Computed Solution: Diffuse part View independant solution Hardware rendering Ray Trace: View dependant part Save image Interactivity: Render - Cache iMAGIS EGWR /06/2000 Results: Quality control 4 sec 1200 particles 5 sec 7800 particles 15 sec particles Vary ct parameter iMAGIS EGWR /06/2000 Indirect 1 min 42 sec 4 min 34 sec iMAGIS EGWR /06/2000 Particle tracing comparison Complex, indirectly lit scene simulation 10 min Particle trace Our method iMAGIS EGWR /06/2000 Video VIDEO iMAGIS EGWR /06/2000 Conclusion Integrated algorithm Hierarchical Radiosity with Clustering and Particle Tracing Guide particle emission with Links Place particles during push-pull Handles indirect light well Rapid computation Interactive simulations for small scenes Fast coarse solutions for complex scenes iMAGIS EGWR /06/2000 Future Work Separate Reconstruction Low and High frequencies Dynamic updates Partial particle shooting Distributed/Monte-Carlo Ray-trace Solution with importance Local precise solution Detect needed interactions iMAGIS EGWR /06/2000 The End