LIGHT TRANSPORT 25/11/2011 Shinji Ogaki
4 Papers Progressive Photon Beams Lightslice: Matrix Slice Sampling for Many- Lights Problem Modular Radiance Transfer Practical Filtering for Efficient Ray-Traced Directional Occlusion
PROGRESSIVE PHOTON BEAMS Wojciech Jarosz et at.
1.Cast Photons 2.Gather Photon Mapping Photon Query Point Fixed Search Radius
LS+DS+E Paths Accurate Caustics Unlimited # of Photons Progressive Photon Mapping Photon Reverse Photon Search Radius
Extension to Volume (LS+MS+E Paths) PPB (Progressive Photon Beam) Photon Beam Query Ray
L: Radiance Tr: Transmittance s: Surface m: Media σs: Scattering Coefficient f: Phase Function Radiative Transport Equation Photon Beam Query Ray Xs S Xw W
Beam x Beam 1D Estimator FluxKernel Scattering Coef
Results
LIGHTSLICE: MATRIX SLICE SAMPLING FOR MANY-LIGHTS PROBLEM Jiawei Ou et al.
Many-Lights Problem Global Illumination (Diffuse Indirect Illum.) Matrix Interpretation of Many-Lights VPL (Virtual Point Light)
Many-Lights Problem Global Illumination (Diffuse Indirect Illum.) Matrix Interpretation of Many-Lights VPL (Virtual Point Light)
Many-Lights Problem Global Illumination (Diffuse Indirect Illum.) Matrix Interpretation of Many-Lights VPL (Virtual Point Light)
Many-Lights Problem Global Illumination (Diffuse Indirect Illum.) Matrix Interpretation of Many-Lights VPL (Virtual Point Light)
Many-Lights Problem Global Illumination (Diffuse Indirect Illum.) Matrix Interpretation of Many-Lights VPL (Virtual Point Light)
Many-Lights Problem Global Illumination (Diffuse Indirect Illum.) Matrix Interpretation of Many-Lights VPL (Virtual Point Light)Sample
Transport Matrix Close to Low Rank..
Algorithm 1.Matrix Slicing 2.Slice Sampling 3.Initial Light Clustering 4.Per Cluster Refinement
Results SliceVisualization
Results (cont’d) Lightslice MRCS Lightcut
Limitations Parameter Selection (# of Slices etc.) Glossy Surface Animation Matrix Sparsity Comprehensive Comparison is missing (Coherent Light Cut and Pixelcuts?)
MODULAR RADIANCE TRANSFER Bradford J. Loos et al.
Module Patched Local is Global Module
Shapes
Transport Matrix (Local) F: Direct to Indirect Transfer (One Bounce) Sample
Reduced Direct-to-Indirect Transfer in Shape Truncated SVD of F Not so Sparse, Unfortunately Sample
Reduced Direct-to-Indirect Transfer in Shape (cont’d) Light Prior (Basis for Plausible Direct Lighting) Id1Id2Idm……
Reduced Direct-to-Indirect Transfer in Shape (cont’d) Truncated SVD of M Very Sparse Sample
Reduced Direct-to-Indirect Transfer between Shapes (Local to Global) Interface
Results
Limitations Lighting Condition outside of the Light Prior High Frequency Glossy Transport Large Scale Indirect Shadows within Blocks Dictionary Shapes (e.g. Internal Occluders) User Interface
PRACTICAL FILTERING FOR EFFICIENT RAY-TRACED DIRECTIONAL OCCLUSION Kevin Egan et al.
Ambient Occlusion ( )/5=0.6 Hemisphere
1.Cast Rays 2.Filter Ambient Occlusion with a Sparse Set of Rays Expensive Cheap
Distant Lighting in Linear Sub-Domains
Frequency Analysis and Sheared Filtering Light(y) Receiver(x) x y Occluder Spectrum Bandlimited by Filter Flatland Scene Occlusion Function f(x, y) 0 Receiver(x) 1 0 Light(y) 1 x y Occluders x y
Frequency Analysis and Sheared Filtering (cont’d)
Rotationally-Invariant Filter Infinitesimal Sub-domains
Results 6+ mins to filter
Limitations Artifacts due to Undersampling in the 1st Pass Smoothes out some Areas of Detail Noise in Areas where Brute Force Computation is used