Reduced Aggregate Jan Novk Scattering Operators Ralf Habel s s - - PowerPoint PPT Presentation

▶
reduced aggregate
SMART_READER_LITE
LIVE PREVIEW

Reduced Aggregate Jan Novk Scattering Operators Ralf Habel s s - - PowerPoint PPT Presentation

Sep 16, 2023 132 likes •643 views

Adrian Blumer Reduced Aggregate Jan Novk Scattering Operators Ralf Habel s s Derek Nowrouzezahrai for Path Tracing Wojciech Jarosz Overview Reduced Aggregate Scattering Operators for Path Tracing 2 Overview Reduced Aggregate

slide-1
SLIDE 1

s

Reduced Aggregate Scattering Operators

for Path Tracing

Adrian Blumer Jan Novák Ralf Habel Derek Nowrouzezahrai Wojciech Jarosz

s

slide-2
SLIDE 2

Overview

Reduced Aggregate Scattering Operators for Path Tracing

2

slide-3
SLIDE 3

Overview

Reduced Aggregate Scattering Operators for Path Tracing

3

slide-4
SLIDE 4

Overview

Reduced Aggregate Scattering Operators for Path Tracing

4

slide-5
SLIDE 5

Overview

Reduced Aggregate Scattering Operators for Path Tracing

5

slide-6
SLIDE 6

Overview

Reduced Aggregate Scattering Operators for Path Tracing

6

slide-7
SLIDE 7

Motivation

image sources: clouds: https://www.flickr.com/photos/rocor/608762581 tree: http://www.wallpapersonview.com/wallpapers/landscape_astonishing_nature_quality_trees_background_picture-2560x1600-12555.html snowmen: http://funmozar.com/christmas-snow-wallpapers/ sheep: https://travelblog.expedia.co.th/westhern_thailand/15043/

7

slide-8
SLIDE 8

Selected Related Work

Clustered principal components for precomputed radiance transfer

Sloan et al., 2003

Modular radiance transfer

Loos et al., 2011

A Practical Model for Subsurface Light Transport

Henrik et al., 2001 Much more: importance sampling, caching of light transport, vegetation rendering, subsurface scattering, …

9

slide-9
SLIDE 9

Precomputation Path tracer integration

slide-10
SLIDE 10

Precomputation Path tracer integration

slide-11
SLIDE 11

Method

Light transport as a matrix operation

transport matrix direct indirect

12

indirect illumination direct illumination

× =

slide-12
SLIDE 12

indirect

Method

Light transport as a matrix operation

basis transform transport basis transform direct

13

indirect illumination direct illumination

= × × ×

slide-13
SLIDE 13

Method

Learning reduced bases

configuration 1

# of vertices

14

Modular radiance transfer Loos et al., 2011

slide-14
SLIDE 14

Method

Learning reduced bases

configuration 1

# of vertices

configuration 2

15

Modular radiance transfer Loos et al., 2011

slide-15
SLIDE 15

Method

Learning reduced bases

configuration 1

# of vertices

configuration 2 ...

16

Modular radiance transfer Loos et al., 2011

slide-16
SLIDE 16

Method

Learning reduced bases

configuration 1

# of vertices

configuration 2 ... ...

17

Modular radiance transfer Loos et al., 2011

slide-17
SLIDE 17

Method

Learning reduced bases

configuration 1

# of vertices

configuration 2 ... ... ...

18

Modular radiance transfer Loos et al., 2011

slide-18
SLIDE 18

Method

Learning reduced bases

# of vertices # of configuarions

19

Modular radiance transfer Loos et al., 2011

slide-19
SLIDE 19

Method

Learning reduced bases

# of vertices # of configurations

Compress using SVD + truncate

20

Modular radiance transfer Loos et al., 2011

slide-20
SLIDE 20

Method

Learning reduced bases

# of vertices # of basis functions

compact basis

21

Modular radiance transfer Loos et al., 2011

slide-21
SLIDE 21

Method

Learning reduced bases

# of vertices # of basis functions

compact basis

22

Modular radiance transfer Loos et al., 2011

slide-22
SLIDE 22

Method

Learning reduced bases

# of vertices # of basis functions

compact basis

23

Modular radiance transfer Loos et al., 2011

slide-23
SLIDE 23

Method

Learning reduced bases

# of vertices # of basis functions

compact basis

24

Modular radiance transfer Loos et al., 2011

slide-24
SLIDE 24

Method

Learning reduced bases

# of vertices # of basis functions

clustered compact basis

25

Clustered principal components for precomputed radiance transfer Sloan et al. 2003

slide-25
SLIDE 25

Method

Learning reduced bases

direct indirect indirect basis transform transport basis transform direct basis transform

26

indirect illumination direct illumination

= × × ×

slide-26
SLIDE 26

Precomputation Path tracer integration

slide-27
SLIDE 27

Path tracer integration

Monte Carlo evaluation

29

slide-28
SLIDE 28

Path tracer integration

batching and caching

30

direct illumination cache

slide-29
SLIDE 29

Path tracer integration

batching and caching

31

direct illumination cache

slide-30
SLIDE 30

Path tracer integration

batching and caching

32

direct illumination cache

slide-31
SLIDE 31

Path tracer integration

batching and caching

33

indirect illumination cache direct illumination cache between batches, apply transport

slide-32
SLIDE 32

Path tracer integration

batching and caching

34

direct illumination cache indirect illumination cache

slide-33
SLIDE 33

Path tracer integration

Importance caching and sampling

35

slide-34
SLIDE 34

Path tracer integration

Importance caching and sampling

36

Importance cache

slide-35
SLIDE 35

Path tracer integration

Importance caching and sampling

37

Importance cache

slide-36
SLIDE 36

Path tracer integration

Importance caching and sampling

38

between batches, apply transport sampling weights Importance cache

slide-37
SLIDE 37

Path tracer integration

Importance caching and sampling

39

sampling weights

slide-38
SLIDE 38

Path tracer integration

Importance caching and sampling

40

sampling weights

slide-39
SLIDE 39

Results

slide-40
SLIDE 40

Quality

Side lighting (in the training set) Path tracing Path tracing with RASO

*indirect illumination only

42

slide-41
SLIDE 41

Quality

Back-lighting (not in the training set) Path tracing Path tracing with RASO

*indirect illumination only

43

slide-42
SLIDE 42

Quality

Ambient sky (not in the training set) Path tracing Path tracing with RASO

*indirect illumination only

44

slide-43
SLIDE 43

Quality

Front-lighting with occluder (not in the training set) Path tracing Path tracing with RASO

*indirect illumination only

45

slide-44
SLIDE 44

Performance

Equal-time Path tracing

46

slide-45
SLIDE 45

Performance

Equal-time Path tracing with RASOs (ours)

47

slide-46
SLIDE 46

TODO: video? Temporal stability

48

slide-47
SLIDE 47

Homogeneous volume

(17 min) (6 s) (3.1 min)

multiple-scattering only, discretization: 1283 voxel grid, dipole parameters hand-tweaked for visually similar result

49

PT with RASO (6 s) PT Diffusion dipole PT with RASO

slide-48
SLIDE 48

Conclusion

Limitations

  • precomputation & storage
  • bias

50

Advantages

  • fast convergence
  • perceptually unobtrusive error
  • temporal stability
slide-49
SLIDE 49

Future Work

  • All-frequency Transport
  • Application to Clouds
  • Non-negative matrix factorization

51

slide-50
SLIDE 50

Thanks for your time! Questions?