A Framework for Transient Rendering Adrian Jarabo 1 Julio Marco 1 - - PowerPoint PPT Presentation

▶
a framework for transient rendering
SMART_READER_LITE
LIVE PREVIEW

A Framework for Transient Rendering Adrian Jarabo 1 Julio Marco 1 - - PowerPoint PPT Presentation

Mar 28, 2023 672 likes •1.35k views

A Framework for Transient Rendering Adrian Jarabo 1 Julio Marco 1 Adolfo Muoz 1 Raul Buisan 1 Wojciech Jarosz 2 Diego Gutierrez 1 1 Universidad de Zaragoza 2 Disney Research Zrich Steady-State Light Transport Infinite Speed of Light

slide-1
SLIDE 1

A Framework for Transient Rendering

Adrian Jarabo1 Julio Marco1 Adolfo Muñoz1 Raul Buisan1 Wojciech Jarosz2 Diego Gutierrez1

2Disney Research Zürich 1Universidad de Zaragoza

slide-2
SLIDE 2

Steady-State Light Transport

Infinite Speed of Light

slide-3
SLIDE 3

Steady-State Light Transport

slide-4
SLIDE 4

Steady-State Light Transport

Infinite Speed of Light

slide-5
SLIDE 5

Transient Light Transport

Finite Speed of Light

slide-6
SLIDE 6

Transient Light Transport

Movie …..…Light Travels (per frame)……. Edgerton’s Stroboscope ……………………………………... Picosecond Resolution ………………………………….

slide-7
SLIDE 7

Transient Light Transport

So if we see at picosecond resolution…

slide-8
SLIDE 8

Transient Light Transport

slide-9
SLIDE 9

Transient Light Transport

But, is breaking this assumption really useful?

slide-10
SLIDE 10

Femto-Photography [Velten2013]

slide-11
SLIDE 11

Femto-Photography [Velten2013]

slide-12
SLIDE 12
  • Visible geometry [Wu2014,OToole2014…]
  • Transparent Objects [Kadambi2013]
  • Hidden geometry [Velten2012…]
  • Reflectance [Naik2011…]
  • GI Components Separation [Wu2014…]
  • …
slide-13
SLIDE 13

Simulation helps:

  • Forward-model for inverse problems
  • Can test new systems before building them
  • Freedom to tweak the physics
slide-14
SLIDE 14

The Problem

slide-15
SLIDE 15

The Problem

slide-16
SLIDE 16

The Problem

slide-17
SLIDE 17

The Problem

slide-18
SLIDE 18

The Problem

Camera Light Participating Media

slide-19
SLIDE 19

The Problem

slide-20
SLIDE 20

The Problem

slide-21
SLIDE 21

The Problem

slide-22
SLIDE 22

The Problem

slide-23
SLIDE 23

The Problem

slide-24
SLIDE 24

The Problem

slide-25
SLIDE 25

The Problem

slide-26
SLIDE 26

The Problem

OK OK Bad Bad Bad

slide-27
SLIDE 27
  • 1. How to reconstruct time-resolved light?
  • 2. How to distribute samples along time?

The Problem

slide-28
SLIDE 28
  • 1. How to reconstruct time-resolved light?
  • 2. How to distribute samples along time?

Our Contribution

slide-29
SLIDE 29
  • 1. How to reconstruct time-resolved light?
  • 2. How to distribute samples along time?

Our Contribution

slide-30
SLIDE 30
slide-31
SLIDE 31

Histogram Density Estimation

[Jarabo2012, OToole2014, Ament2014]

Reconstructed Signal

slide-32
SLIDE 32

Kernel-Based Density Estimation

Reconstructed Signal

slide-33
SLIDE 33

Kernel-Based Density Estimation Progressive

slide-34
SLIDE 34

Kernel-Based Density Estimation Progressive

Iteration i Iteration i-1

slide-35
SLIDE 35

Kernel-Based Density Estimation Progressive

slide-36
SLIDE 36

Binning Kernel-Based

slide-37
SLIDE 37
slide-38
SLIDE 38
  • 1. How to reconstruct time-resolved light?
  • 2. How to distribute samples along time?

Our Contribution

slide-39
SLIDE 39
  • 1. How to reconstruct time-resolved light?
  • 2. How to distribute samples along time?

Our Contribution

slide-40
SLIDE 40

OK OK Bad Bad Bad

slide-41
SLIDE 41
slide-42
SLIDE 42

Time-Based Sampling

slide-43
SLIDE 43

Time-Based Sampling

slide-44
SLIDE 44

Time-Based Sampling

slide-45
SLIDE 45
slide-46
SLIDE 46
slide-47
SLIDE 47

Set of techniques for time-based sampling in participating media 1. Next Segment Distance 2. Shadow Connection 3. Angular Sampling

Time-Sampling

slide-48
SLIDE 48
slide-49
SLIDE 49
slide-50
SLIDE 50
slide-51
SLIDE 51

Kernel-Based Density Estimation + Time Sampling

slide-52
SLIDE 52
slide-53
SLIDE 53
  • 1. How to reconstruct time-resolved light
  • 2. How to distribute samples along time

Our Contribution

slide-54
SLIDE 54

Additional Results

slide-55
SLIDE 55
slide-56
SLIDE 56

Including: 1. Birefringency 2. Chromatic dispersion in time 3. Comparison with captured data

More Results in the Supplementary Video

slide-57
SLIDE 57
  • Error introduced by Kernel DE

Signal-aware Kernel Bandwidth [Kaplanyan2013] Error Metric [Hachisuka2010]

  • Sampling Surface Light Transport

Caustic in time → Manifold Exploration [Jakob2012]

Discussion & Future Work

slide-58
SLIDE 58
  • Help developing new techniques using

transient light propagation

  • Educational tool
  • Useful for other fields?

– Astrophysics, Neutron Transport, Sound Rendering….

Discussion & Future Work

slide-59
SLIDE 59
  • 1. Formalized Transient Rendering
  • 2. Kernel-Based Reconstruction for Transient LT
  • 3. Sampling Techniques along Time
  • 4. Non-trivial effects of Transient LT

Code, Videos and Data at:

http://giga.cps.unizar.es/~ajarabo/pubs/transientSIGA14

Conclusions

slide-60
SLIDE 60
slide-61
SLIDE 61
slide-62
SLIDE 62

Set of techniques for time-based sampling in participating media 1. Next Segment Distance 2. Shadow Connection 3. Angular Sampling

Time-Sampling

slide-63
SLIDE 63
  • 1. Next subpath Segment Distance
slide-64
SLIDE 64
  • 2. Shadow Connection
slide-65
SLIDE 65
  • 3. Angular Sampling
slide-66
SLIDE 66
  • Rad. Sampling

Histogram Time Sampling Histogram

  • Rad. Sampling

Kernel-DE Time Sampling Kernel-DE