Post on 03-Jan-2016
transcript
Advected texturesAdvected textures
Fabrice Neyret
EVASION - GRAVIR / IMAG - INRIA
Grenoble, France
Purpose: amplifying fluid simulationPurpose: amplifying fluid simulation
Purpose: amplifying fluid simulationPurpose: amplifying fluid simulation
Why not simply increase resolution ?Why not simply increase resolution ?
Cost: N log(N) / time step with N=1000^3
Storage
Problems with CFD for CG [Lamorlette&Foster 02]
Unknown small scale phenomena vs artist desires + phenomenological knowledge
Why not simply increase resolution ?Why not simply increase resolution ?
Cost: N log(N) / time step with N=1000^3
Storage
Problems with CFD for CG [Lamorlette&Foster 02]
Unknown small scale phenomena vs artist desires + phenomenological knowledge
Why not simply increase resolution ?Why not simply increase resolution ?
Cost: N log(N) / time step with N=1000^3
Storage
Problems with CFD for CG [Lamorlette&Foster 02]
Unknown small scale phenomena vs artist desires + phenomenological knowledge
Why not simply increase resolution ?Why not simply increase resolution ?
Cost: N log(N) / time step with N=1000^3
Storage
Problems with CFD for CG [Lamorlette&Foster 02]
Unknown small scale phenomena vs artist desires + phenomenological knowledge
Advecting texturesAdvecting textures
= Advecting u,v [Max&Becker 96, Stam 99]
Regeneration blending 3 dephased textures
(illusion of motion)
latency = life duration
Problems with texture advectionProblems with texture advection
Choosing the latency
Blending textures
Sub-animation
11
22
33
Problems with texture advectionProblems with texture advection
Choosing the latency
Blending textures
Sub-animation
11
22
33
Problems with texture advectionProblems with texture advection
Choosing the latency
Blending textures ghosting effects
Sub-animation
11
22
33
Problems with texture advectionProblems with texture advection
Choosing the latency
Blending textures
Sub-animation
11
22
33
Problems with texture advectionProblems with texture advection
Choosing the latency
Blending textures
Sub-animation
11
22
33
1.1. Advecting textures1.1. Advecting textures
A latency value is ok for a range of velocities (V) bad motion illusion if V < texture stretching if V >
1. Advecting textures1. Advecting textures
Adapting latency locally Layers of given latency + masks
Local criterion
– cumulated deform = particle integral of ||
– Target deform d*
1. Advecting textures1. Advecting textures
Adapting latency locally Layers of given latency + masks
Local criterion
– cumulated deform = particle integral of ||
– Target deform d*
Layer 1: lat1
Layer 2: lat2 (>lat1)
Layer 3: lat3 (>lat2)
1. Advecting textures1. Advecting textures
Adapting latency locally Layers of given latency + masks
Local criterion
– cumulated deform = particle integral of ||
– Target deform d*
1 2 3
2.2. Blending textures2.2. Blending textures
Image textures
Procedural textures
2. Blending textures: image textures2. Blending textures: image textures
What to do ? (morphing ?)
2. Blending textures: procedural textures2. Blending textures: procedural textures
Our solution:
3.3. Sub-animation3.3. Sub-animation
Flownoise [Perlin&Neyret 01]
3. Sub-animation3. Sub-animation
Flownoise for sub-scales
rotations vorticity spectrum
Kolmogorov cascade
3. Sub-animation3. Sub-animation
Flownoise for sub-scales
rotations vorticity spectrum < >
Kolmogorov cascade
3. Sub-animation3. Sub-animation
Flownoise for sub-scales
rotations vorticity spectrum < >
Kolmogorov cascade
k = k =
EE
kk00 kk
sub-gridsub-gridCFDCFD
microscalemicroscale
3. Sub-animation3. Sub-animation
Vorticity energy transfer through scales
distribution law for ( power law ) only needs to scale it ( estimate )
Our case:
heterogeneous fluid ( locality no Fourier ) not at equilibrium ( transfer delay time ) user control
relaxation
< >< >
3. Sub-animation3. Sub-animation
Vorticity energy transfer through scales
distribution law for ( power law ) only needs to scale it ( estimate )
Our case:
heterogeneous fluid ( locality no Fourier ) not at equilibrium ( transfer delay time ) user control
relaxation
(user-defined parameters k and k )
< >< >
ResultsResults
ConclusionConclusion
Mixing lo-res CFD and hi-res animated texture: A model efficient & controllable
Future work:– 3D applications: detailed clouds & avalanches– Better flownoise control– Manage empty space– Hardware procedural shader
Advected texturesAdvected textures
Fabrice Neyret
EVASION - GRAVIR / IMAG - INRIA
(Grenoble, France)
Sub-animation parametersSub-animation parameters
parameters k and k:
small : reactive high : inertial
small : viscous high : light
1. Advecting textures1. Advecting textures
3D 3D