Download presentation
Presentation is loading. Please wait.
Published byShanon Pope Modified over 9 years ago
1
1
2
SIGGRAPH 2010 Single Scattering in Heterogeneous Participating media Cyril Delalandre Pascal Gautron Jean-Eudes MarvieGuillaume François Technicolor Research & Innovation The Moving Picture Company 2
3
Real Light Scattering Light interaction with heterogeneous participating media 3
4
Participating Media Characterized by: –σ a : absorption coefficient –σ s : scattering coefficient –σ t : σ a + σ s, extinction coefficient –D(s) : medium density at the point s 4
5
Single Scattering in participating media 5
6
Explicit ray-marching KiKi P in P out K in K out P3P3 P2P2 P1P1 PnPn 6
7
Outline Introduction Previous works Our algorithm Results Conclusion 7
8
Outline Introduction Previous worksPrevious works Our algorithm Results Conclusion 8
9
Previous works [Zhou08] [Gautron09] [Jansen10] 9
10
Volumetric Shadow Mapping [Gautron09], Siggraph Talk 2009 +Real-time rendering -Homogeneous medium 10
11
Real-time smoke rendering by using compensed ray marching, [Zhou08], Siggraph 2008 +Real-Time rendering +Single and Multiple scattering +Heterogeneous medium -Heavy precomputation 11
12
Fourier Opacity Map, [Jansen10], I3D 2010 +Real-time rendering +Heterogeneous medium -No scattering computation 12
13
Outline Introduction Previous works Our algorithmOur algorithm Results Conclusion 13
14
Goals Single scattering Volume shadows Dynamic Viewpoint and Lighting Generic and Dynamic Medium Scalable Without pre-computation 14
15
Acceleration of the light reduced intensity computation Small memory footprint Contributions 15 Attenuation Function Map
16
Algorithm Light Depth Map Attenuation function map View Depth & radiance Map Scattering Computation 16
17
Attenuation Function Map C 0 = 0 C 1 = 0. C n = 0 C 0 += att 1 x C 1 += att 1 x. C n += att 1 x C 0 += att 2 x C 1 += att 2 x. C n += att 2 x C 0 += att n x C 1 += att n x. C n += att n x Light Point of View 1.Opaque objects Depth Map 2.For each pixel: a.Bounding box Intersection b.Coefficient Computation by ray- marching 17
18
Attenuation Function Map Light Point of View 1.Opaque objects Depth Map 2.For each pixel: a.Bounding box Intersection b.Coefficient Computation by ray- marching c.Store in Multi Render Targets C 0-n (0,0) C 0-n (K,0)C 0-n (K,L) C 0-n (0,L) C 0-n (i,j) 18
19
Scene Rendering In User point of view: 1.Compute the depth and reflected radiance of opaque objects (R,G,B)(Alpha) 19
20
Compute Scattering C 0-n (0,0) C 0-n (K,0 ) C 0-n (K,L) C 0-n (0,L) C 0-n (i,j ) 2.Intersection of the medium bounding box 20 Depth Map Attenuation Function Map
21
Compute Scattering C 0-n (0,0) C 0-n (K,0 ) C 0-n (K,L) C 0-n (0,L) C 0-n (i,j ) 3.Perform a ray-marching 21 Depth Map Attenuation Function Map
22
Compute Scattering C 0-n (0,0) C 0-n (K,0 ) C 0-n (K,L) C 0-n (0,L) C 0-n (i,j ) For each ray sample: a.Check the sample visibility b.Check the light visibility c.Lri computation using the coefficients Lri = c 0 x+ c 1 x + … + c n x 22 Depth Map Attenuation Function Map
23
Results Comparison between our algorithm and an explicit ray-marching algorithm 100 samples per ray / 16 DCT coefficients 23 Speed-up 30x
24
Ringing artifacts High Density Medium 24
25
Add a scale factor to avoid ringing artifacts High Density Medium Transformed Signal Reconstructed Signal 25
26
High Density Medium Without scale factorWith scale factor 26 Explicit ray-marching
27
Demos – GTX 480 – 720p 27 500 samples / view ray, 16 DCT coefficients, 256 3 medium size, 10 FPS
28
Demos – GTX 480 – 720p 28 100 samples / view ray, 16 DCT coefficients, 128 3 medium size, 19 FPS
29
Conclusion 29 Single Scattering Volume Shadows Fast Lri computation Scalable Generic No pre-computation
30
Large medium Multiple Scattering Particle representation Image Lighting Future Works 30
31
Thanks to A. W. Bargteil and C. Thompson (University of Utah) for the fluid simulation Thanks to K. Bouatouch for his research supervision Acknowledgements 31 Questions ?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.