Progressive Volume Rendering of Large Unstructured Grids Steven P. - - PowerPoint PPT Presentation

▶
progressive volume rendering of large unstructured grids
SMART_READER_LITE
LIVE PREVIEW

Progressive Volume Rendering of Large Unstructured Grids Steven P. - - PowerPoint PPT Presentation

Feb 21, 2023 582 likes •863 views

Progressive Volume Rendering of Large Unstructured Grids Steven P. Callahan 1 , Louis Bavoil 1 , Valerio Pascucci 2 , and Cludio T. Silva 1 1 SCI Institute, University of Utah 2 Lawrence Livermore National Laboratory Motivation Large-scale

slide-1
SLIDE 1

Progressive Volume Rendering of Large Unstructured Grids

Steven P. Callahan1, Louis Bavoil1, Valerio Pascucci2, and Cláudio T. Silva1

1 SCI Institute, University of Utah 2 Lawrence Livermore National Laboratory

slide-2
SLIDE 2

Motivation

 Large-scale simulations produce a lot of data  Interactive visualization techniques not keeping up  Meshes may be too large to render locally

2

slide-3
SLIDE 3

Progressive Volume Rendering

3% 0.01 sec 33% 7 sec 66% 18 sec 100% 34 sec

3

slide-4
SLIDE 4

Objective

 Progressive Rendering

  • Show intermediate results
  • Reuse intermediate results
  • Allow user interrupt
  • Only render pertinent data

 Client-Server Architecture

  • Support a thin client with limited memory
  • Standard server used as a data repository
  • Facilitate remote visualization

4

slide-5
SLIDE 5

Issues

 Tetrahedra are not natively supported

  • Projected Tetrahedra
  • [Shirley and Tuchman ‘90, Wiley et al. ‘02]

 Compositing requires strict order

  • Visibility Sorting
  • [Williams et al. ‘92]
  • Ray Casting
  • [Bunyk et al. ‘97, Weiler et al. ‘03]
  • Hybrids
  • [Farias et al. ‘00, Callahan et al. ‘05]

5

slide-6
SLIDE 6

Issues

 Hierarchical level-of-detail not suitable

  • Regular Sampling
  • [Leven et al. 2002]
  • Geometry Simplification
  • [Cignoni et al. 2005]
  • LOD Without Hierarchies
  • [Callahan et al. 2005]

 Remote Visualization difficult using a standard server

  • Image Transmission
  • [Engel et al. 2000]
  • Uncomposited Image Transmission
  • [Bethel et al. 2000]
  • Data Transmission
  • [Lippert et al. 1997, Engel et al. 1998, Kaehler et al. 2004]

6

slide-7
SLIDE 7

Background

 Hardware-Assisted Visibility Sorting

  • Sort in both object-space and image-space

CPU GPU [Callahan et al. 2005] http://havs.sourceforge.net and vtk/ParaView

7

slide-8
SLIDE 8

Background

 Dynamic Level-of-Detail 2.0 fps 5.3 fps 10.0 fps 16.1 fps [Callahan et al. 2005] http://havs.sourceforge.net and vtk/ParaView

8

slide-9
SLIDE 9

Overview

 Server: Processes geometry and transmits triangles in visibility order  Client: Receives geometry and renders it progressively

9

  • ฀

฀ ฀ ฀ ฀ ฀

slide-10
SLIDE 10

 Preprocess

  • Create min-max octree

 Geometry Server  Octree Traversal  Object-Space Sort

The Server

10

  • ฀

฀ ฀ ฀ ฀ ฀

slide-11
SLIDE 11

The Server

 Preprocess  Geometry Server

  • Calculate depth range

 Octree Traversal  Object-Space Sort

11

  • ฀

฀ ฀ ฀ ฀ ฀

slide-12
SLIDE 12

The Server

 Preprocess  Geometry Server  Octree Traversal

  • Cull range geometry
  • Frustum cull geometry

 Object-Space Sort

12

  • ฀

฀ ฀ ฀ ฀ ฀

slide-13
SLIDE 13

The Server

 Preprocess  Geometry Server  Octree Traversal  Object-Space Sort

  • Sort geometry by centroid
  • Compress and send

13

  • ฀

฀ ฀ ฀ ฀ ฀

slide-14
SLIDE 14

The Client

 Preprocess

  • Get boundary geometry from

server

  • Build pre-integration table

 Interactive Mode  Progressive Mode  Completed Mode

14

  • ฀

฀ ฀ ฀ ฀ ฀

slide-15
SLIDE 15

The Client

 Preprocess  Interactive Mode

  • Volume render the boundary

geometry

  • Keep the back boundary fragments

 Progressive Mode  Completed Mode

15

  • ฀

฀ ฀ ฀ ฀ ฀

slide-16
SLIDE 16

The Client

 Preprocess  Interactive Mode  Progressive Mode

  • Render range of geometry
  • Display progressive image

 Completed Mode

16

  • ฀

฀ ฀ ฀ ฀ ฀

  • zbegin

zend

slide-17
SLIDE 17

 Use three buffers to render progressive image

  • Complete: finished volume rendering
  • Active: temporary storage of k fragments
  • Progressive: Complete blended with approximation

Progressive Mode

17

Complete Active Progressive

slide-18
SLIDE 18

 Pass 1:

  • Render geometry into Active buffer
  • Composite overflow fragments into Complete buffer.

Progressive Mode

Complete Active

18

slide-19
SLIDE 19

Progressive Mode

 Pass 2:

  • Render empty space into Progressive buffer using

Active buffer and back boundary fragments

19

Approximate Empty Space

slide-20
SLIDE 20

Progressive Mode

 Pass 3:

  • Composite Complete buffer into Progressive buffer
  • Display Progressive buffer
  • Keep Complete and Active buffers for next progressive

step Approximate = Complete + Approximate

20

slide-21
SLIDE 21

Overview

 Preprocess  Interactive Mode  Progressive Mode  Completed Mode

  • Composite Active buffer into

Complete buffer

  • Display and store Complete buffer

21

  • ฀

฀ ฀ ฀ ฀ ฀

slide-22
SLIDE 22

Results

22

slide-23
SLIDE 23

Results

23

slide-24
SLIDE 24

Considerations

 The network  Transfer functions  Other interaction methods

24

slide-25
SLIDE 25

Conclusion

 Remote visualization of large unstructured grids  Progressions converge to full-quality renderings  Allows interactive exploration of large datasets  Future Work:

  • Cutting planes
  • Stream compression
  • Time-varying data

25

slide-26
SLIDE 26

Acknowledgments

 Carlos Scheidegger, Huy Vo  Datasets

  • Neely and Batina (NASA)
  • O’Hallaran and Shewchuck (CMU)

 Funding

  • DOE
  • IBM
  • SNL
  • LLNL
  • ARO
  • University of Utah

26