Ray Tracing Animations Using 4D Kd-Trees Jens Olsson Examensarbete - - PowerPoint PPT Presentation

▶

ray tracing animations using 4d kd trees

Ray Tracing Animations Using 4D Kd-Trees Jens Olsson Examensarbete - - PowerPoint PPT Presentation

Jan 30, 2023 208 likes •954 views

Ray Tracing Animations Using 4D Kd-Trees Jens Olsson Examensarbete 2007-02-19 Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 1 / 29 Outline Outline Introduction Ray-Tracing Motion Blur Acceleration

slide-1

SLIDE 1

Ray Tracing Animations Using 4D Kd-Trees

Jens Olsson Examensarbete 2007-02-19

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 1 / 29

slide-2

SLIDE 2

Outline

Outline

Introduction

Ray-Tracing Motion Blur Acceleration Structures

Project

Motivation Theory Implementation

Results

Data Conclusions Future Work

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 2 / 29

slide-3

SLIDE 3

Outline

Outline

Introduction

Ray-Tracing Motion Blur Acceleration Structures

Project

Motivation Theory Implementation

Results

Data Conclusions Future Work

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 2 / 29

slide-4

SLIDE 4

Outline

Outline

Introduction

Ray-Tracing Motion Blur Acceleration Structures

Project

Motivation Theory Implementation

Results

Data Conclusions Future Work

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 2 / 29

slide-5

SLIDE 5

Outline

Outline

Introduction

Ray-Tracing Motion Blur Acceleration Structures

Project

Motivation Theory Implementation

Results

Data Conclusions Future Work

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 2 / 29

slide-6

SLIDE 6

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Ray-Tracing

Technique used to generate convincing imagery Mimics a camera, but reversed light flow Millions of rays are traced through the scene Very general, extends naturally from simple to complex models

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 3 / 29

slide-7

SLIDE 7

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Ray-Tracing

Technique used to generate convincing imagery Mimics a camera, but reversed light flow Millions of rays are traced through the scene Very general, extends naturally from simple to complex models

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 3 / 29

slide-8

SLIDE 8

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Ray-Tracing

Technique used to generate convincing imagery Mimics a camera, but reversed light flow Millions of rays are traced through the scene Very general, extends naturally from simple to complex models

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 3 / 29

slide-9

SLIDE 9

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Ray-Tracing

Technique used to generate convincing imagery Mimics a camera, but reversed light flow Millions of rays are traced through the scene Very general, extends naturally from simple to complex models

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 3 / 29

slide-10

SLIDE 10

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Ray-Tracing

Technique used to generate convincing imagery Mimics a camera, but reversed light flow Millions of rays are traced through the scene Very general, extends naturally from simple to complex models

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 3 / 29

slide-11

SLIDE 11

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Motion Blur

Camera film is exposed for a short moment Moving objects exposure are spread out over the film The objects appear as smeared in the accumulated image

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 4 / 29

slide-12

SLIDE 12

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Motion Blur

Camera film is exposed for a short moment Moving objects exposure are spread out over the film The objects appear as smeared in the accumulated image

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 4 / 29

slide-13

SLIDE 13

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Motion Blur

Camera film is exposed for a short moment Moving objects exposure are spread out over the film The objects appear as smeared in the accumulated image

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 4 / 29

slide-14

SLIDE 14

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Motion Blur

Camera film is exposed for a short moment Moving objects exposure are spread out over the film The objects appear as smeared in the accumulated image

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 4 / 29

slide-15

SLIDE 15

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Motion Blur contd.

Simulated by stochastic sampling, accumulation buffers or blur post-process

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 5 / 29

slide-16

SLIDE 16

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-17

SLIDE 17

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-18

SLIDE 18

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-19

SLIDE 19

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-20

SLIDE 20

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-21

SLIDE 21

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-22

SLIDE 22

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-23

SLIDE 23

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-24

SLIDE 24

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-25

SLIDE 25

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-26

SLIDE 26

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Acceleration Structures

Ray tracing mainly performs visibility queries A strategy is needed to make visibility queries fast All strategies pre-sort scene primitives in some way Grids , Bounding Volume Hierarchies , BSP Trees , Ray Classification etc etc etc

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 6 / 29

slide-27

SLIDE 27

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Kd-Trees

Simple data structure in the BSP tree family Scene volume is recursively subdivided by arbitrarily positioned axis-aligned splitting planes Much research has already been done, regarded as one of the most efficient acceleration structures Existing algorithms for construction as well as traversal

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 7 / 29

slide-28

SLIDE 28

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Kd-Trees

Simple data structure in the BSP tree family Scene volume is recursively subdivided by arbitrarily positioned axis-aligned splitting planes Much research has already been done, regarded as one of the most efficient acceleration structures Existing algorithms for construction as well as traversal

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 7 / 29

slide-29

SLIDE 29

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Kd-Trees

Simple data structure in the BSP tree family Scene volume is recursively subdivided by arbitrarily positioned axis-aligned splitting planes Much research has already been done, regarded as one of the most efficient acceleration structures Existing algorithms for construction as well as traversal

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 7 / 29

slide-30

SLIDE 30

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Kd-Trees

Simple data structure in the BSP tree family Scene volume is recursively subdivided by arbitrarily positioned axis-aligned splitting planes Much research has already been done, regarded as one of the most efficient acceleration structures Existing algorithms for construction as well as traversal

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 7 / 29

slide-31

SLIDE 31

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Kd-Trees

Simple data structure in the BSP tree family Scene volume is recursively subdivided by arbitrarily positioned axis-aligned splitting planes Much research has already been done, regarded as one of the most efficient acceleration structures Existing algorithms for construction as well as traversal

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 7 / 29

slide-32

SLIDE 32

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Kd-Trees contd.

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 8 / 29

slide-33

SLIDE 33

Intro Project Results Ray-Tracing Motion Blur Acceleration Structures

Kd-Trees contd.

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 9 / 29

slide-34

SLIDE 34

Intro Project Results Motivation Theory Implementation

Motivation

Easy to formulate time-dependent ray tracing Attractive properties for rendering sequences of images Practical problems when actually implementing it Efficient techniques lays the foundation for more advanced future work (ie. global illumination)

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 10 / 29

slide-35

SLIDE 35

Intro Project Results Motivation Theory Implementation

Motivation

Easy to formulate time-dependent ray tracing Attractive properties for rendering sequences of images Practical problems when actually implementing it Efficient techniques lays the foundation for more advanced future work (ie. global illumination)

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 10 / 29

slide-36

SLIDE 36

Intro Project Results Motivation Theory Implementation

Motivation

Easy to formulate time-dependent ray tracing Attractive properties for rendering sequences of images Practical problems when actually implementing it Efficient techniques lays the foundation for more advanced future work (ie. global illumination)

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 10 / 29

slide-37

SLIDE 37

Intro Project Results Motivation Theory Implementation

Motivation

Easy to formulate time-dependent ray tracing Attractive properties for rendering sequences of images Practical problems when actually implementing it Efficient techniques lays the foundation for more advanced future work (ie. global illumination)

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 10 / 29

slide-38

SLIDE 38

Intro Project Results Motivation Theory Implementation

Motivation

Easy to formulate time-dependent ray tracing Attractive properties for rendering sequences of images Practical problems when actually implementing it Efficient techniques lays the foundation for more advanced future work (ie. global illumination)

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 10 / 29

slide-39

SLIDE 39

Intro Project Results Motivation Theory Implementation

Theory

We wish to adapt kd-trees to store time-dependent data We need to subdivide primitives temporally We need a new heuristic for constructing the tree

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 11 / 29

slide-40

SLIDE 40

Intro Project Results Motivation Theory Implementation

Theory

We wish to adapt kd-trees to store time-dependent data We need to subdivide primitives temporally We need a new heuristic for constructing the tree

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 11 / 29

slide-41

SLIDE 41

Intro Project Results Motivation Theory Implementation

Theory

We wish to adapt kd-trees to store time-dependent data We need to subdivide primitives temporally We need a new heuristic for constructing the tree

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 11 / 29

slide-42

SLIDE 42

Intro Project Results Motivation Theory Implementation

Theory contd.

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 12 / 29

slide-43

SLIDE 43

Intro Project Results Motivation Theory Implementation

Theory contd.

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 13 / 29

slide-44

SLIDE 44

Intro Project Results Motivation Theory Implementation

Theory contd.

Temporal segmentation rule find the largest tb > ta such that S

B
∪iB(Pti)
≤ κ S(B(Pta)) , ti ∈ [ta, tb]

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 14 / 29

slide-45

SLIDE 45

Intro Project Results Motivation Theory Implementation

Theory contd.

Nodes are split using a heuristic Most popular heuristic for kd-trees are Surface Area Heuristics (SAH) Makes a local estimate of the future costs of traversal and intersection tests of the resulting subnodes

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 15 / 29

slide-46

SLIDE 46

Intro Project Results Motivation Theory Implementation

Theory contd.

Nodes are split using a heuristic Most popular heuristic for kd-trees are Surface Area Heuristics (SAH) Makes a local estimate of the future costs of traversal and intersection tests of the resulting subnodes

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 15 / 29

slide-47

SLIDE 47

Intro Project Results Motivation Theory Implementation

Theory contd.

Nodes are split using a heuristic Most popular heuristic for kd-trees are Surface Area Heuristics (SAH) Makes a local estimate of the future costs of traversal and intersection tests of the resulting subnodes

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 15 / 29

slide-48

SLIDE 48

Intro Project Results Motivation Theory Implementation

Theory contd.

The SAH cost function C = ct + (1 − be)(pANAci + pBNBci) We reinterpreted these for the temporal case

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 16 / 29

slide-49

SLIDE 49

Intro Project Results Motivation Theory Implementation

Theory contd.

The SAH cost function C = ct + (1 − be)(pANAci + pBNBci) We reinterpreted these for the temporal case

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 16 / 29

slide-50

SLIDE 50

Intro Project Results Motivation Theory Implementation

Theory contd.

Spatial case pA = p(BA|B) = SBA SB pB = p(BB|B) = SBB SB Temporal case pA = p(BA|B) = |tBA| |tB| pB = p(BB|B) = |tBB| |tB|

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 17 / 29

slide-51

SLIDE 51

Intro Project Results Motivation Theory Implementation

Theory contd.

Spatial case pA = p(BA|B) = SBA SB pB = p(BB|B) = SBB SB Temporal case pA = p(BA|B) = |tBA| |tB| pB = p(BB|B) = |tBB| |tB|

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 17 / 29

slide-52

SLIDE 52

Intro Project Results Motivation Theory Implementation

Implementation

Developed a simple ray tracer Written in C++ Plug-in inside off-the-shelf 3D software

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 18 / 29

slide-53

SLIDE 53

Intro Project Results Motivation Theory Implementation

Implementation

Developed a simple ray tracer Written in C++ Plug-in inside off-the-shelf 3D software

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 18 / 29

slide-54

SLIDE 54

Intro Project Results Motivation Theory Implementation

Implementation

Developed a simple ray tracer Written in C++ Plug-in inside off-the-shelf 3D software

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 18 / 29

slide-55

SLIDE 55

Intro Project Results Data Conclusions Future Work

Data

The 3 test scenes

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 19 / 29

slide-56

SLIDE 56

Intro Project Results Data Conclusions Future Work

Data contd.

Head Test

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 20 / 29

slide-57

SLIDE 57

Intro Project Results Data Conclusions Future Work

Data contd.

Head Test Results Head test common data method accelerator tree memory size primitives memory size accumulation 3D kd-tree 0.2 MB 0.9 MB stochastic 4D kd-tree 18.4 MB 12.2 MB

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 21 / 29

slide-58

SLIDE 58

Intro Project Results Data Conclusions Future Work

Data contd.

Head test results accumulation 8 subframes 16 subframes 32 subframes 64 subframes time (min) 1.29 2.63 4.99 8.87 total rays 1961956 3909952 7806048 15599764 stochastic 4 samples 8 samples 16 samples 32 samples time (min) 1.25 1.95 3.37 6.22 total rays 933204 1886628 3822216 7687188

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 22 / 29

slide-59

SLIDE 59

Intro Project Results Data Conclusions Future Work

Data contd.

Head test rendering time breakdown (seconds) accumulation 8 subframes 16 subframes 32 subframes 64 subframes update data 4.3 8.3 16.3 26.1 build tree 2.6 5.5 10.3 17.6 raytrace 70.0 143.5 271.6 487.6 stochastic 4 samples 8 samples 16 samples 32 samples update data 1.5 1.5 1.5 1.5 build tree 26.6 26.4 26.0 26.0 raytrace 45.8 87.9 173.7 344.8

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 23 / 29

slide-60

SLIDE 60

Intro Project Results Data Conclusions Future Work

Data contd.

Stochastic rendering time comparison (minutes) kd-tree segmented 4 samples 8 samples 16 samples 32 samples 3D no 16.7 36.3 65.4 130.7 3D yes 2.3 3.5 6.1 11.1 4D yes 1.2 1.9 3.4 6.2 Stochastic rendering time breakdown (seconds) kd-tree segmented update data build tree 3D no 1.2 0.5 3D yes 1.5 51.9 4D yes 1.5 26.0

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 24 / 29

slide-61

SLIDE 61

Intro Project Results Data Conclusions Future Work

Conclusions

Possible to store animation in a 4D kd-tree Better performance compared to storing the animation in a 3D kd-tree Memory requirements were higher than expected Interesting applications for 4D kd-trees

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 25 / 29

slide-62

SLIDE 62

Intro Project Results Data Conclusions Future Work

Conclusions

Possible to store animation in a 4D kd-tree Better performance compared to storing the animation in a 3D kd-tree Memory requirements were higher than expected Interesting applications for 4D kd-trees

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 25 / 29

slide-63

SLIDE 63

Intro Project Results Data Conclusions Future Work

Conclusions

Possible to store animation in a 4D kd-tree Better performance compared to storing the animation in a 3D kd-tree Memory requirements were higher than expected Interesting applications for 4D kd-trees

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 25 / 29

slide-64

SLIDE 64

Intro Project Results Data Conclusions Future Work

Conclusions

Possible to store animation in a 4D kd-tree Better performance compared to storing the animation in a 3D kd-tree Memory requirements were higher than expected Interesting applications for 4D kd-trees

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 25 / 29

slide-65

SLIDE 65

Intro Project Results Data Conclusions Future Work

Conclusions

Possible to store animation in a 4D kd-tree Better performance compared to storing the animation in a 3D kd-tree Memory requirements were higher than expected Interesting applications for 4D kd-trees

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 25 / 29

slide-66

SLIDE 66

Intro Project Results Data Conclusions Future Work

Applications - Temporally instanced primitives

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 26 / 29

slide-67

SLIDE 67

Intro Project Results Data Conclusions Future Work

Applications - Temporally instanced primitives

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 27 / 29

slide-68

SLIDE 68

Intro Project Results Data Conclusions Future Work

Applications - Temporally instanced primitives

10 10

1

10

2

10

3

10

4

10

5

10

6

10

5

10

6

10

7

10

8

10

9

10

10

10

11

10

12

memory usage number of primitives memory load / byte no instancing temporal instancing sampled data extrapolated data

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 28 / 29

slide-69

SLIDE 69

Intro Project Results Data Conclusions Future Work

Future Work

Triangle interpolation needs to be way faster; is it possible to cpu- and cache-optimize the triangle interpolation in an efficient way? Costs have to be calculated at least twice for every node; Any way of avoiding this without explicitly relating temporal and spatial axes? Global Illumination for animations; How can we take advantage of a time-dependent structure? Investigate applications further; Temporal instances could be used for crowd simulations

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 29 / 29

slide-70

SLIDE 70

Intro Project Results Data Conclusions Future Work

Future Work

Triangle interpolation needs to be way faster; is it possible to cpu- and cache-optimize the triangle interpolation in an efficient way? Costs have to be calculated at least twice for every node; Any way of avoiding this without explicitly relating temporal and spatial axes? Global Illumination for animations; How can we take advantage of a time-dependent structure? Investigate applications further; Temporal instances could be used for crowd simulations

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 29 / 29

slide-71

SLIDE 71

Intro Project Results Data Conclusions Future Work

Future Work

Triangle interpolation needs to be way faster; is it possible to cpu- and cache-optimize the triangle interpolation in an efficient way? Costs have to be calculated at least twice for every node; Any way of avoiding this without explicitly relating temporal and spatial axes? Global Illumination for animations; How can we take advantage of a time-dependent structure? Investigate applications further; Temporal instances could be used for crowd simulations

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 29 / 29

slide-72

SLIDE 72

Intro Project Results Data Conclusions Future Work

Future Work

Triangle interpolation needs to be way faster; is it possible to cpu- and cache-optimize the triangle interpolation in an efficient way? Costs have to be calculated at least twice for every node; Any way of avoiding this without explicitly relating temporal and spatial axes? Global Illumination for animations; How can we take advantage of a time-dependent structure? Investigate applications further; Temporal instances could be used for crowd simulations

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 29 / 29

slide-73

SLIDE 73

Intro Project Results Data Conclusions Future Work

Future Work

Triangle interpolation needs to be way faster; is it possible to cpu- and cache-optimize the triangle interpolation in an efficient way? Costs have to be calculated at least twice for every node; Any way of avoiding this without explicitly relating temporal and spatial axes? Global Illumination for animations; How can we take advantage of a time-dependent structure? Investigate applications further; Temporal instances could be used for crowd simulations

Jens Olsson Ray Tracing Animations Using 4D Kd-Trees Examensarbete 2007-02-19 29 / 29