Download Virtual Spherical Lights for Many-Light - Cornell University PDF

TitleVirtual Spherical Lights for Many-Light - Cornell University
LanguageEnglish
File Size4.7 MB
Total Pages49
Table of Contents
                            Effects of Global Illumination Approximations on Material Appearance
Global illumination rendering
Global illumination rendering
Overview
Related work
Related work – Visual equivalence
Related work – VPL rendering
Related work – VPL rendering
VPL rendering is fast, but…
VPLs for high-fidelity rendering
VPL Rendering Parameters
VPL Rendering Parameters
Space of rendering parameters
Psychophysical experiments
Test objects – Shape complexity
Test objects – Materials
Test objects – Materials
Test objects – Materials
Test objects – Materials
Scene
Stimulus images
Stimulus images –VPL count
Stimulus images – Clamping level
Experiment 1: Artifact visibility
Experiment 2: Material change
Data analysis
Experiment results
Trends – VPL count
Trends – VPL count
Trends – VPL count
Trends – VPL count
Trends – VPL count
Trends – Material contrast gloss
Trends – Shape complexity
Trends – Shape complexity
Trends – Illumination
Validation
Applications
Application – Per-object clamping
Summary of results
Conclusions
Future work
Acknowledgements
Thank you
Additional Slides
Ambiguity: highlights  vs. artifacts
Apps: Luminance normalization
Apps: Luminance normalization
Trends – Material roughness
                        
Document Text Contents
Page 1

Jaroslav
Křivánek

James
Ferwerda

Kavita
Bala

Effects of Global Illumination
Approximations on Material

Appearance

Cornell University &
Charles University, Prague

Rochester Institute of
Technology

Cornell University

Presenter
Presentation Notes
In this work we investigate how some approximations to GI affect image fidelity and material appearance.

Page 2

• Required for accurate appearance, but slow

2

Global illumination rendering

scene: Autodesk | rendering: Edgar Velázquez-Armendáriz

Presenter
Presentation Notes
GI is necessary for accurate appearance rendering , but it’s slow.

Page 24

• Standard two-alternative forced choice method
• 480 trials, 12 participants 24

Experiment 1: Artifact visibility

cl
a

m
pi

n
g

le
ve

l

VPL count

Presenter
Presentation Notes
So now we’ve defined the stimulus images. The next step are the perceptual experiments themselves.

In the first experiment, the goal is to identify the parameter combinations that produce visible artifacts.

For each trial, we display two images: One is a path traced reference (with no artifacts) and the other VPL rendering (possibly with artifacts).
The subjects are asked to select the image that has the artifacts.

Page 25

• Standard two-alternative forced choice method
• 520 trials, 14 participants 25

Experiment 2: Material change

VPL count

cl
a

m
pi

n
g

le
ve

l

Presenter
Presentation Notes
Similarly in the Material change experiment, the goal is to identify the parameter combination that produce material changes.

In each trial, we show the “reference” on the top. At the bottom, there is a pair of images, one of them path traced and the second rendered with VPLs. We ask the subjects to select which of the two objects at the bottom has different material from the reference on the top.

Page 48

48

Apps: Luminance normalization

Page 49

49

Trends – Material roughness

Similer Documents