Realtime realistic illumination and shading of stratiform clouds

Real-time realistic illumination and shading of stratiform clouds Antoine Bouthors, Fabrice Neyret, Sylvain Lefebvre Evasion-GRAVIR / IMAG-INRIA Grenoble, France Eurographics Workshop on Natural Phenomena

Motivations l Clouds are very complex: Complex shape l Lots of visual features l Lots of details l l No one has reproduced everything yet

Cloud facts

Clouds types

Multiple scattering No absorption, albedo = 1 l Thick (up to several km) l Dense (Mean free path = several meters) → Very high number of scattering events l (up to several hundreds)

Mie phase function Strongly anisotropic l Computationally expensive l • Depends on droplet size, temperature, wavelength 49% 1%

Droplet size distribution l One common phase function for the whole cloud

Visual features l Fogbow & glory

Visual features Anisotropic forward scattering l Diffusive multiple scattering l

Visual features l Pseudo-specular effect

Visual features l Clouds-ground inter-reflections Ice blink Water sky

Previous work Offline approaches Musgrave et. al. Musgrave

Previous work Interactive approaches l Different approximations: Low albedo / low density → single scattering l Simpler phase function → not all features l Diffusion approximation → isotropic l l Volumetric models l l Very coarse Billboards or slices l Lots of overdraw Harris et. al.

Our Model

Overview l What we assume: Stratiform → locally equivalent to a slab l Homogeneous density l l What we want: Realistic l Real-time (GPU-friendly) l Allow animation l

Overview l Complex behavior l Rays with different orders of scattering l Our idea: a study of the contribution of each order

Overview l Our analysis: l Narrow forward scattering only important on silhouettes (low orders) Low orders of scattering give anisotropic features (fogbow, glory, pseudo-specular) l Higher orders become diffusive (i. e. isotropic) l Sky and ground play a role l l Sky illuminance = ~25% of that of the sun l Standard floor reflectance = ~20%

Overview l Our analysis: l Narrow forward scattering only important on silhouettes (low orders) Low orders of scattering give anisotropic features (fogbow, glory, pseudo-specular) l Higher orders become diffusive (i. e. isotropic) l Sky and ground play a role l l Sky illuminance = ~25% of that of the sun l Standard floor reflectance = ~20%

Overview l Our analysis: l Narrow forward scattering only important on silhouettes (low orders) Low orders of scattering give anisotropic features (fogbow, glory, pseudo-specular) l Higher orders become diffusive (i. e. isotropic) l Sky and ground play a role l l Sky illuminance = ~25% of that of the sun l Standard floor reflectance = ~20%

Overview l Our analysis: l Narrow forward scattering only important on silhouettes (low orders) Low orders of scattering give anisotropic features (fogbow, glory, pseudo-specular) l Higher orders become diffusive (i. e. isotropic) l Sky and ground play a role l l Sky illuminance = ~25% of that of the sun l Standard floor reflectance = ~20%

Overview l Scattering: Strong forward scattering: special treatment l 1+2 orders: analytic l Higher orders (3+) considered isotropic l l Environment: Take sky into account (diffuse source) l Clouds-ground radiosity (plane parallel) l l Shape: l Height field

Clouds representation l Clouds are stored as a height field Well fitted for stratiform clouds l Lot of details in a small space l Procedural & animatable l

Phase function l Our Modified-Mie model Strong narrow (<5°) forward scattering moved into extinction function l Error negligible with multiple scattering l

Phase function l Modified-Mie model validation Monte-Carlo bench: generated reflection BRDFs 7+ 65 4 7+ 3 1 2 4 6 5 3 2 1

Single scattering l Slab -> Local analytical function

Double scattering l Local approximation by convolution

3+ scattering Convolution = bad idea for high orders l Assume slab → 1 D problem → column l

3+ scattering Assuming diffusion → simple interaction between column cells l Given scattering behavior of a cell → analytical solution for the column l

3+ scattering l How to know the multiple scattering behavior of one cell ? Characteristic of the « cloud material » l Precompute Monte Carlo integration → reflectance & transmittance of a cell l

3+ scattering l Issue: l l Diffuse hypothesis broken on top cell: anisotropy still plays a role Solution: 1 D model correction term dependant on the sun incident angle l Correction parameters fitted on Monte Carlo simulations l

Sky & ground contributions Previous: valid for directional L, V l Sky & ground: Assumed diffuse source l 1 D model used l → gives diffuse reflectance & transmittance l

Ground-clouds inter-reflections l l l Radiosity between two facing parallel planes → known form factors (analytic) Heterogeneous source (clear sky + cloud bottom) Heterogenous reflectance (ground AND clouds) Ecloud = (Sun+sky)T + Σi(Rcloud_i FF Eground_i) Eground = (Sun+sky)R + Σi(Rground_i FF Ecloud_i)

Ground-clouds inter-reflections Reflectances and radiosities in textures l Plane-plane → Form Factor (ring to d. S) l l → Using MIP-mapping = l - GPU-enhanced iterative algorithm Render-to-texture l Hardware MIP-mapping l

Validation l Validation of our 1 D model

Validation l Comparison of our model with a Monte. Carlo integration 5 m-thick slab Monte-Carlo bench: generated reflection BRDFs 3 1 2 100 m-thick slab 3 2 1

Implementation l Height field: advected textures [Ney 03] 16 km-wide landscape l 2 km-height, 500 m-thick cloud layer l l Shaders + radiosity on GPU 512 x 512 clouds textures l 512 x 512 ground textures l 512 x 512 shadows textures l 16 x 16 radiosity textures l l 18 to 40 FPS on current hardware

Results

Lighting contributions bottom view 1 and 2 scattering 3+ scattering Ground illumination (incl. radiosity) Sky illumination

Summing it all bottom view

Lighting contributions top view 1 and 2 scattering 3+ scattering Ground illumination Sky illumination

Summing it all top view

Features Glory and fogbow

Features Glory and fogbow

Features Water sky

Features Pseudo-specular reflection

Features Pseudo-specular reflection

Videos l Videos

Conclusion l Good Realistic l Reproduce all clouds visual features l Account for clouds-ground inter-reflections l Real-time l Animation-friendly l l Bad Poor lateral shading (clouds borders) l 3+ scattering can be improved l Limited to stratiform clouds l

Questions ?