Tsuyoshi Yamamoto
-
Upload
nyssa-garrison -
Category
Documents
-
view
36 -
download
0
description
Transcript of Tsuyoshi Yamamoto
Hokkaido University
Efficient Rendering of Lightning Efficient Rendering of Lightning Taking into Account Scattering Taking into Account Scattering
Effects due to Clouds and Effects due to Clouds and Atmospheric ParticlesAtmospheric Particles
Tsuyoshi YamamotoTsuyoshi Yamamoto
Tomoyuki NishitaTomoyuki Nishita(The University of Tokyo)
Yoshinori DobashiYoshinori Dobashi
(Hokkaido University)
Hokkaido University
OverviewOverview
ā¢ Introductionā¢ Introduction
ā¢ Effects of Atmospheric Scattering due to Lightningā¢ Effects of Atmospheric Scattering due to Lightning
ā¢ Clouds Illuminated by Lightningā¢ Clouds Illuminated by Lightning
ā¢ Resultsā¢ Results
ā¢ Conclusionsā¢ Conclusions
Hokkaido University
IntroductionIntroductionā¢ Photo-realistic Rendering of Natural Scenesā¢ Photo-realistic Rendering of Natural Scenes
- visual assessments- flight simulators
- visual assessments- flight simulators
Previous methods:Clear/cloudy days
ā¢ Simulations under bad weather conditionsā¢ Simulations under bad weather conditions
- windstorm, sandstorm, rain, lightning- windstorm, sandstorm, rain, lightning
- Realistic rendering of scenes including
lightning
- Development of efficient rendering method
ā¢ Purposesā¢ Purposes
Hokkaido University
Important ElementsImportant Elements
ā¢ Shape of lightningā¢ Shape of lightning ā¢ Illuminating cloudsā¢ Illuminating clouds
ā¢ Atmospheric scatteringā¢ Atmospheric scatteringā¢ Illuminating groundā¢ Illuminating ground
Hokkaido University
Previous MethodsPrevious Methods
ā¢ Methods related to lightningā¢ Methods related to lightning
Modeling using particle systems [Reed94]
Probabilistic modeling [Kruszewski99]
DLA taking into account clouds [Sosorbaram01]
cloudsatmospheric
scatteringrendering
speed
-
-
-
-
- slow
fast
slow
Hokkaido University
Previous MethodsPrevious Methods
Modeling using particle systems [Reed94]
Probabilistic modeling [Kruszewski99]
DLA taking into account clouds [Sosorbaram01]
cloudsatmospheric
scatteringrendering
speed
-
-
-
-
- slow
fast
slow
ā¢ Methods related to lightningā¢ Methods related to lightning
Hokkaido University
Previous MethodsPrevious Methods
Modeling using particle systems [Reed94]
Probabilistic modeling [Kruszewski99]
DLA taking into account clouds [Sosorbaram01]
cloudsatmospheric
scatteringrendering
speed
-
-
-
-
- slow
fast
slow
ā¢ Methods related to lightningā¢ Methods related to lightning
Hokkaido University
Previous MethodsPrevious Methods
Modeling using particle systems [Reed94]
Probabilistic modeling [Kruszewski99]
DLA taking into account clouds [Sosorbaram01]
cloudsatmospheric
scatteringrendering
speed
-
-
-
-
- slow
fast
slow
ā¢ Methods related to lightningā¢ Methods related to lightning
fast Our method
Hokkaido University
hardware rendering of smokes
shafts of light through trees and clouds[Max86]
[Dobashi00]hardware rendering of shafts of light through clouds
[Jansen98]photon map[Rushmeier87]extending radiosity method[Nishita87]shafts of light produced by spotlights
[Stam99, Stam01]
Previous MethodsPrevious Methodsā¢ Related to clouds and atmospheric scatteringā¢ Related to clouds and atmospheric scattering
None of these takes into account scattering effects due to lightning flash.
None of these takes into account scattering effects due to lightning flash.
Hokkaido University
ā¢ Use of Reedās method for modelingā¢ Use of Reedās method for modeling
ā¢ User specifies Color of lightningā¢ User specifies Color of lightning
ā¢ Atmospheric scattering due to flash of lightningā¢ Atmospheric scattering due to flash of lightning
ā¢ Clouds illuminated by flash of lightningā¢ Clouds illuminated by flash of lightning
ā¢ Efficient rendering of clouds for fly- through animationsā¢ Efficient rendering of clouds for fly- through animations
Features of Proposed MethodFeatures of Proposed Method
ā¢ Use of Reedās method for modelingā¢ Use of Reedās method for modeling
ā¢ User specifies Color of lightningā¢ User specifies Color of lightning
ā¢ Atmospheric scattering due to flash of lightningā¢ Atmospheric scattering due to flash of lightning
ā¢ Clouds illuminated by flash of lightningā¢ Clouds illuminated by flash of lightning
ā¢ Efficient rendering of clouds for fly- through animationsā¢ Efficient rendering of clouds for fly- through animations
Hokkaido University
ā¢ Use of Reedās method for modelingā¢ Use of Reedās method for modeling
ā¢ User specifies Color of lightningā¢ User specifies Color of lightning
ā¢ Atmospheric scattering due to flash of lightningā¢ Atmospheric scattering due to flash of lightning
ā¢ Clouds illuminated by flash of lightningā¢ Clouds illuminated by flash of lightning
ā¢ Efficient rendering of clouds for fly- through animationsā¢ Efficient rendering of clouds for fly- through animations
Features of Proposed MethodFeatures of Proposed Method
Hokkaido University
OverviewOverview
ā¢ Introductionā¢ Introduction
ā¢ Effects of Atmospheric Scattering due to Lightningā¢ Effects of Atmospheric Scattering due to Lightning
ā¢ Clouds Illuminated by Lightningā¢ Clouds Illuminated by Lightning
ā¢ Resultsā¢ Results
ā¢ Conclusionsā¢ Conclusions
Hokkaido University
Atmospheric ScatteringAtmospheric Scattering
viewpoint
clouds
lightning
point sources
ā¢ Placing point light sources
Hokkaido University
viewpoint V
P
ā¢ Consider a single source k
s
t
Ik
dtIs
tsFI k
aaa
keye )(
))(exp()(cos)(
20
)(
a: density
F : phase function: phase anglea: extinction coefficient
s : distance between source and P
t : distance between V and P
Ik : intensity of point source
: wavelength
Atmospheric ScatteringAtmospheric Scatteringā¢ Placing point light sources
Hokkaido University
viewpoint
P
s
t
Ik
dtIs
tsFI k
aaa
keye )(
))(exp()(cos)(
20
)(
ā¢ No analytical solutions
Atmospheric ScatteringAtmospheric Scattering
ā¢ Consider a single source k
ā¢ Placing point light sources
Hokkaido University
dtIs
tsFI k
aaa
keye )(
))(exp()(cos)(
20
)(
ā¢ No analytical solutions
Atmospheric ScatteringAtmospheric Scattering
ā¢ Consider a single source k
ā¢ Placing point light sources
viewpoint
Ik
Ieye(k)
ā¢ Ray tracing
ā¢ Computationally expensiveā¢ Computationally expensive
ā¢ Use of look-up tableā¢ Use of look-up table
Hokkaido University
Efficient Computation Using Look-up TableEfficient Computation Using Look-up Table
ā¢ Creating look-up table
ā¢ Intensity due to a single sourcedtI
s
tsFI k
aaa
keye )(
))(exp()(cos)(
20
)(
s
t
(ueye, veye)
P
(u, v)- local coordinate uv
),,()()( eyeeyelkk
eye vuIII
22eyevus || eyeuut
22/1cos eyevu
duvu
uuvu
vuFvuI
eye
eyeeyeaa
ueye
aeyeeyeleye
22
22
22
|))|(exp(
)1
(),,(
u
v
Hokkaido University
ā¢ function of (ueye, veye , )
ā¢ preparing table by changing values of (ueye, veye , )
ā¢ -T < (ueye, veye) < Tć (T:specified by user)ā¢ : sampled at R, G, B
Efficient Computation Using Look-up TableEfficient Computation Using Look-up Table
s
t
(ueye, veye)
P
(u, v)
u
v
),,()()( eyeeyelkk
eye vuIII
Hokkaido University
Efficient Computation Using Look-up TableEfficient Computation Using Look-up Table
n
kkpkplkp lvuIII
1,, ),,()(
ā¢ Intensity of pixel
Can be computed efficientlyusing look-up table
ā¢ function of (ueye, veye , )
ā¢ preparing table by changing values of (ueye, veye , )
s
t
(ueye, veye)
P
(u, v)
u
v
),,()()( eyeeyelkk
eye vuIII
Hokkaido University
OverviewOverview
ā¢ Introductionā¢ Introduction
ā¢ Effects of Atmospheric Scattering due to Lightningā¢ Effects of Atmospheric Scattering due to Lightning
ā¢ Clouds Illuminated by Lightningā¢ Clouds Illuminated by Lightning
ā¢ Resultsā¢ Results
ā¢ Conclusionsā¢ Conclusions
Hokkaido University
- voxels
Intensity Calculation of CloudsIntensity Calculation of Cloudsā¢ Density distributionā¢ Density distribution
- metaballs
[Dobashi00]metaballs
R
q
effectiveradius
center density
metaball
field function
Hokkaido University
pointsources
lightning
- voxels
Intensity Calculation of CloudsIntensity Calculation of Cloudsā¢ Density distributionā¢ Density distribution
- metaballs
[Dobashi00]metaball
- use of LOD
- use of hardware
ā¢ Intensity Calculationā¢ Intensity Calculation
- sum of intensity due to each point source
Hokkaido University
pointsources
lightning
- 3D voxels
Intensity Calculation of CloudsIntensity Calculation of Cloudsā¢ Density distributionā¢ Density distribution
- metaballs
[Dobashi00]metaball
- use of LOD
- use of hardware
ā¢ Intensity Calculationā¢ Intensity Calculation
- sum of intensity due to each point source
Hokkaido University
Computing Attenuation Using HardwareComputing Attenuation Using Hardware
ā¢ Attenuation to each metaballā¢ Attenuation to each metaball- use of hardware-accelerated splatting [Dobashi00]
- limited to parallel sources- limited to parallel sources
- extending to point sources- extending to point sources
metaball
lightning
pointsource k
Hokkaido University
box as sixscreens
metaball- placing a box as 6 screens
Computing Attenuation Using HardwareComputing Attenuation Using Hardware
ā¢ Attenuation to each metaballā¢ Attenuation to each metaball
pointsource k
Hokkaido University
box as sixscreens
metaball- placing a box as 6 screens
Computing Attenuation Using HardwareComputing Attenuation Using Hardware
ā¢ Attenuation to each metaballā¢ Attenuation to each metaball
pointsource k
Hokkaido University
box as sixscreens
- place billboards at centers of metaballs
billboard(square polygon)
- project metaballs
- pixel value of the centers
attenuation ratio
ā¢ Problemā¢ Problem
cost no. of metaballsācost no. of metaballsā(realistic clouds ļ¼ tens of thousands of metaballs)
Using LODUsing LOD ļ¼ļ¼grouping metaballs grouping metaballs hierarchicallyhierarchically
Using LODUsing LOD ļ¼ļ¼grouping metaballs grouping metaballs hierarchicallyhierarchically
- placing a box as 6 screens
Computing Attenuation Using HardwareComputing Attenuation Using Hardware
ā¢ Attenuation to each metaballā¢ Attenuation to each metaball
pointsource k
Hokkaido University
Efficient Computation Using LODEfficient Computation Using LOD
metaballj
Ik
Ikj
r
ā¢ Light reaching metaball ļ¼ā¢ Light reaching metaball ļ¼
- inversely proportional to square of distance- inversely proportional to square of distance
pointsource k
22
))))((exp(exp())(())((
rr
rrIIII kk
kjkj
Hokkaido University
- attenuation due to cloud particles- attenuation due to cloud particles
metaball j
Ik
Ikj
r
pointsource k
Efficient Computation Using LODEfficient Computation Using LOD
ā¢ Light reaching metaball ļ¼ā¢ Light reaching metaball ļ¼
- inversely proportional to square of distance- inversely proportional to square of distance
22
))))((exp(exp())(())((
rr
rrIIII kk
kjkj
Hokkaido University
metaball
- intensity is small at distant regions, and almost uniform.- intensity is small at distant regions, and almost uniform.
Efficient Computation Using LODEfficient Computation Using LOD
ā¢ Light reaching metaball ļ¼ā¢ Light reaching metaball ļ¼
pointsource k
- attenuation due to cloud particles- attenuation due to cloud particles
- inversely proportional to square of distance- inversely proportional to square of distance
22
))))((exp(exp())(())((
rr
rrIIII kk
kjkj
Hokkaido University
metaball
ā¢ Approximation by larger metaballsā¢ Approximation by larger metaballsā¢ Selecting appropriate metaballs depending on distancesā¢ Selecting appropriate metaballs depending on distances
Efficient Computation Using LODEfficient Computation Using LOD
ā¢ Light reaching metaball ļ¼ā¢ Light reaching metaball ļ¼
pointsource k
- intensity is small at distant regions, and almost uniform.- intensity is small at distant regions, and almost uniform.
- attenuation due to cloud particles- attenuation due to cloud particles
- inversely proportional to square of distance- inversely proportional to square of distance
22
))))((exp(exp())(())((
rr
rrIIII kk
kjkj
Hokkaido University
ā¢ Representing metaballs using octreeā¢ Representing metaballs using octree
ā¢ Selecting appropriate levels depending on distances
ā¢ Selecting appropriate levels depending on distances
metaball
larger metaballā¢ Grouping neighboring metaballsā¢ Grouping neighboring metaballs
- density ļ¼ average- radius ļ¼ twice- density ļ¼ average- radius ļ¼ twice
Efficient Computation Using LODEfficient Computation Using LOD
Hokkaido University
ā¢ Selection of appropriate levelsā¢ Selection of appropriate levels- energy received by metaball j
- condition ļ¼
( : threshold)
metaball j
point source k Ik
Ikj
r
Ej = (light reaching metaball) x (volume)dVj
Efficient Computation Using LODEfficient Computation Using LOD
dVj2
))(exp()(
r
rIk
2
))(exp()(max
r
rIkdVj
requires integration of density of cloud particlesrequires integration of density of cloud particles
Hokkaido University
ā“
exp(-(r)) < 1.0( )
2
)(max
r
IkdVj
ā¢ Selection of appropriate levelsā¢ Selection of appropriate levels- energy received by metaball j metaball j
point source k Ik
Ikj
r
dVj
Efficient Computation Using LODEfficient Computation Using LOD
- condition ļ¼
energy when there are no particles between metaball and point source.energy when there are no particles between metaball and point source.
2
))(exp()(max
r
rIkdVj
Ej = (light reaching metaball) x (volume)
dVj2
))(exp()(
r
rIk
Hokkaido University
clouds
point source k
2
)}(max{
r
Ik
dVj- condition ļ¼- check metaballs of highest level
ā¢ Selection of appropriate levelsā¢ Selection of appropriate levels
Efficient Computation Using LODEfficient Computation Using LOD
Hokkaido University
- proceed to metaballs of lower levels
ā
Ć
2
)}(max{
r
Ik
dVj- condition ļ¼- check metaballs of highest level
ā¢ Selection of appropriate levelsā¢ Selection of appropriate levels
Efficient Computation Using LODEfficient Computation Using LOD
clouds
point source k
Hokkaido University
Ć
ā
- proceed to metaballs of lower levels
2
)}(max{
r
Ik
dVj- condition ļ¼- check metaballs of highest level
ā¢ Selection of appropriate levelsā¢ Selection of appropriate levels
Efficient Computation Using LODEfficient Computation Using LOD
clouds
point source k
Hokkaido University
Reducing number of metaballs to be processedReducing number of metaballs to be processed
- proceed to metaballs of lower levels
2
)}(max{
r
Ik
dVj- condition ļ¼- check metaballs of highest level
ā¢ Selection of appropriate levelsā¢ Selection of appropriate levels
Efficient Computation Using LODEfficient Computation Using LOD
clouds
point source k
selected metaballs
Hokkaido University
OverviewOverviewā¢ Introductionā¢ Introduction
ā¢ Features of Our methodā¢ Features of Our method
ā¢ Effects of Atmospheric Scattering due to Lightningā¢ Effects of Atmospheric Scattering due to Lightning
ā¢ Clouds Illuminated by Lightningā¢ Clouds Illuminated by Lightning
ā¢ Resultsā¢ Results
ā¢ Conclusionsā¢ Conclusions
Hokkaido University
ResultsResults
ā¢ Verification using simple exampleā¢ Verification using simple example
density of atmospheric particles ļ¼ 0.15density of atmospheric particles ļ¼ 0.15
ā¢ Parameter settings:ā¢ Parameter settings:
attenuation ratio ļ¼ 0.03attenuation ratio ļ¼ 0.03
threshold ļ¼ 0.2threshold ļ¼ 0.2
no. of point sources ļ¼ 50no. of point sources ļ¼ 50
table size ļ¼ 128x128table size ļ¼ 128x128(T: 1.5 [km])(T: 1.5 [km])
no. of metaballs: 250,000no. of metaballs: 250,000
Hokkaido University
ResultsResults
50 times faster!50 times faster!
with LOD without LOD
ā¢ Computation timeā¢ Computation time- with LOD : 8 [s]
- without LOD : 400 [s]
computer ļ¼ PentiumIII (733MHz), GeForce2GTSImage size: 720X480
Hokkaido University
ResultsResults
(a) lightning in clouds (b) multiple lightning
(c) colored lightning (pink) (d) lightning at sunset
Hokkaido University
Example Example AnimationAnimation ļ¼ļ¼ VIDEO)VIDEO)
ā¢ On animating lightning:ā¢ On animating lightning:
ā¢ Simulation of lightning under different conditionsā¢ Simulation of lightning under different conditions
ā¢ Flight simulationā¢ Flight simulation
- Initial points are determined randomly in clouds.
- Periods from occurrence to the extinction are determined randomly, less than 0.5 seconds.
- Intensity is determined randomly.
Hokkaido University
ConclusionsConclusions
ā¢ Realistic image synthesis of scenes including lightningā¢ Realistic image synthesis of scenes including lightning
- atmospheric scattering due to flash of lightning- clouds illuminated by flash of lightning
- efficient rendering using look-up table and idea of LOD
- hierarchical imposters for efficient rendering of clouds
Hokkaido University
Future WorkFuture Work
ā¢ Further speeding up for real-time simulationsā¢ Further speeding up for real-time simulations
ā¢ Automatic determination of parametersā¢ Automatic determination of parameters
ā¢ Automatic determination of lightning colorā¢ Automatic determination of lightning color