Surface Signals for Graphics - PowerPoint PPT Presentation

1 / 20
About This Presentation
Title:

Surface Signals for Graphics

Description:

self-transfer texture (spherical harmonic coefs) Simplicity of regular 2D image ... Jim Blinn, Conal Elliot, Brian Guenter, Hugues Hoppe, Charles Loop, Don Mitchell, ... – PowerPoint PPT presentation

Number of Views:103
Avg rating:3.0/5.0
Slides: 21
Provided by: johns70
Category:

less

Transcript and Presenter's Notes

Title: Surface Signals for Graphics


1
Surface Signals for Graphics
John Snyder Researcher 3D Graphics
Group Microsoft Research
2
Why Surface Signals?
  • Many useful types of surface signals
  • texture map Catmull74, BlinnNewell76 (color)
  • bump map Max81 (normal)
  • displacement map Cook 84 (geometric
    offset)
  • geometry image (geometry)
  • bidirectional texture function (precomputed
    shading)
  • self-transfer texture (spherical harmonic
    coefs)
  • Simplicity of regular 2D image
  • Support on current graphics hardware (e.g. pixel
    shaders)
  • Research questions
  • How to generate and manipulate signals?
  • What new graphics architectures?

3
Surface Signal Research Projects
Creation precomputed radiance transfer
Parameterization signal-specialized param.
Rendering signal-based graphics architecture
Representation geometry images
4
Motivation for Precomputed Transfer
  • better light integration and light transport
  • dynamic, area lights
  • shadowing
  • interreflections
  • in real-time

point light
area light
area lighting, no shadows
area lighting, shadows
5
Self-Transfer Signal (25D)
illuminate
result
Reduces shading to a 25D dot product
(low-frequency lighting)
6
Self-Transfer Results (Diffuse)
No Shadows/Inter Shadows
ShadowsInter
7
Self-Transfer Results (Glossy)
No Shadows/Inter Shadows
ShadowsInter
8
Self-Transfer Demo
9
Parameterization of Surface Signals
Geometry-based (know geometry only)
Signal-specialized (know geometrysignal)
10
Measuring Parameterization Quality
2D texture domain
surface in 3D
11
Geometric Stretch Metric
2D texture domain
surface in 3D
geometric stretch ?2 G2
Parameterize minimize surface integral of
geometric stretch
12
Signal Stretch Metric
domain
surface
f
g
h g?f
signal
Parameterize minimize surface integral of
signal stretch
  • geometric stretch ?f2 Gf2
  • signal stretch ?h2 Gh2

13
Conformal Floater97
Geometric stretch Sander01
Signal stretch Sander02
14
(64x64 texture)
Results Scanned Color
Geometric stretch
Signal stretch
15
Results Normal Map
Geometric stretch
Signal stretch
128x128 texture - multichart
16
Results Precomputed Radiance Transfer
Geometric stretch
Signal stretch
25D signal 256x256 texture
17
3D graphics 2D image processing?
not quite ? use images but of surface signals not
views
  • synthesize images from 3D surface descriptions
  • run-time flexibility change view, lighting,
    rendering params
  • compactness single surface parameterization,
    not multiple views
  • high quality (global illumination) resolution
    independence
  • cheap creation no costly rigs operator, easy
    to edit
  • as preprocess, convert 3D descriptions to 2D
    image reps (surface signals) to accelerate
    run-time
  • signals can be represented as regular 2D images
  • rendering via general, programmable image
    processing ops

18
Rendering Factorization
global illumination computation is too expensive
from scratch
Preprocess(slow)
Run-Time(fast)
surface signals
  • 3D surfaces (meshes)
  • 3D graphics
  • ray tracing, Monte Carlo integration, dynamics
    simulation, encoding
  • 2D images, streams
  • 2D image processing
  • decoding, interpolation / decimation,
    programmable pixel shaders, sample gather

19
End
20
People
Microsoft Research 3D Graphics Group Jim Blinn,
Conal Elliot, Brian Guenter, Hugues Hoppe,
Charles Loop, Don Mitchell, Kirk Olynyk,
Peter-Pike Sloan, John Snyder, Turner Whitted
Collaborators Steven Gortler, Xianfeng Gu, Ziyad
Hakura, Jesse Hall, Jan Kautz, Leonard McMillan,
Pedro Sander, Zoe Wood
Write a Comment
User Comments (0)
About PowerShow.com