High Quality Final Gathering for Hierarchical Monte Carlo Radiosity in General Environments

1
High Quality Final Gathering for Hierarchical
Monte Carlo Radiosityin General Environments
Frederic PérezIgnacio MartínXavier Pueyo
GGG-IIiA/UdGGirona, Spain
2
Outline

• Subject of this work
• Previous work
• Our approach
• Results
• Conclusions
• Current and future work

3
Subject of this work

• Objective
• Rendering of high quality (HQ) images
• For general environments
• Possibly including participating media
• General optical properties
• Accounting for the global illumination (GI)

• Strategy Two-pass method
• 1. Obtain coarse representation of GI
• 2. Refine the solution to get a HQ image

4
Previous Work

• Chen et al. 91
• Progressive Multi-Pass
• Jensen 95
• Importance Driven Photon Maps
• Szirmay-Kalos et al. 98
• Importance Driven Quasi-Random Walk
• Stürzlinger 96, Ureña Torres 97
• Final Gathering

5
Overview of the Algorithm

• 1st Pass Extended HMCR
• Bekaert et al.s Hierarchical Monte Carlo
  Radiosity
• Coarse solution not for rendering

• 2nd Pass Ray Tracing Final Gather
• Improved Importance Sampling with Link
  Probabilities
• Use links as a representation of part of the
  irradiance

6
First Pass Coarse Global Illumination
Extended HMCR
Source
Diffuse surface
Specular surface

• Iterate distributing energy

Medium
7
First Pass Example

• Simple room (direct indirect illumination)

Model
8
Second Pass High Quality Image
RayTracer with Final Gather

• Estimate eye radiance using
• obtained estimations

9
Second Pass Link Probabilities
Importance Sampling with LPs

• estimate incoming light

10
Second Pass Example

• Simple room (direct indirect illumination)

11
Second Pass Noise

• Simple room Progressivity

12
Results

• Office room indirect illumination

13
Results

• Kitchen scene (direct illumination)
• Lambertian light sources (windows)

14
Results

• Kitchen scene More views

15
Conclusions

• Extension of HMCR
• Integration of participating media
• (Glossy diffuse) surfaces
• Final Gathering scheme for the computation of
  high quality images
• Based on the results of the HMCR extended
  algorithm
• Using adaptive Importance Sampling
• Quality of the images does not strongly depend on
  the lighting conditions

16
Current Work

• Natural Lighting
• Sun and sky light integrated in the HMCR
  algorithm

17
Current Work

• Cheaper Rendering Step
• Progressive Radiance Computation Coherence

18
Future Work

• Improve Final Gather
• Scheel et al.s Grid Based Final Gather

19
Acknowledgments

• ESPRIT Open LTR project 35772 SIMULGEN
• http//iiia.udg.es/Simulgen
• Generalitat de Catalunyas 2001/SGR/00296
• CICYTs TIC2001-2932-C03-01

20
High Quality Final Gathering for Hierarchical
Monte Carlo Radiosityin General Environments
Frederic PérezIgnacio MartínXavier Pueyo
frederic_at_ima.udg.eshttp//ima.udg.es/7Efrederic
GGG-IIiA/UdGGirona, Spain
21
Extra slides
22
Second Pass Establishing Links

• Two Possibilities

• Store links during the HMCR step
• All points within a leaf element share set of
  links

• Establish link set during second step
• Each gathering point has its own set of links

23
Application

• Underground station

24
Application

• Underground station

25
First Pass

• HMCR
• Interaction between a surface and a volume

26
First Pass

• HMCR Isotropic media and 3d-textures

Direct viewing vs. Interpolation
Showing the subdivision or not
27
Importance Sampling

• Variance reduction technique
• Use the best samples to evaluate an integral
  Probability Density Function (PDF) prop. to
  kernel
• Reflectance equation
• Importance sampling
• Use of approximate for PDF ?
  Usually by means of a 1st pass

28
Second Pass Link Probabilities

• PDF accuracy and visibility

Solution Adaptive PDFs
29
Results

• Simple scene with participating medium

30
Results

• Another scene with participating medium

31
Results

• Another scene with participating medium