docPatric Ljung Markus Hadwiger Timo Ropinski Christof Rezk
download 
Report Transcription
Patric Ljung Markus Hadwiger Timo Ropinski Christof Rezk
Markus Hadwiger Patric Ljung VR VIS Research Center Vienna, Austria Siemens Corporate p Research Princeton, NJ, USA Christof Rezk Salama Timo Ropinski Computer Graphics Group Institute for Vision and Graphics University of Siegen, Germany Visualization and Computer Graphics Research Group, University of Münster, Germany Ambient Occlusion in DVR Markus Hadwiger Patric Ljung VR VIS Research Center Vienna, Austria Siemens Corporate p Research Princeton, NJ, USA Christof Rezk Salama Timo Ropinski Computer Graphics Group Institute for Vision and Graphics University of Siegen, Germany Visualization and Computer Graphics Research Group, University of Münster, Germany Outline Introduction and Concepts Vicinity Shading Dynamic y Ambient Occlusion Local Ambient Occlusion Global Light Propagation Summary PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 3 Motivation Increased perception of Shapes Densities Depth Issues with surface-based shading Noisy data have poorly defined normals/gradients Surface shading for volumetric objects??? PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 4 Ambient Occlusion Global Light Source I Isotropic i IIncident id Li h Light Visible Hemisphere Integrate Illumination ll i i for a point on the surface Sampled directions or projection of occluders Accuracy: number of / directions and/or resolution of light “cache” PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 5 Vicinity y Shading g A. James Stewart, IEEE Vis 2003 D iffuse Surface Shading Vicinity Shading PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 6 Vicinity y Shading g Concepts p Considers only iso-surface rendering Does not consider transparency Precomputes visibility for all iso-values Each processed direction traverses each voxel exactly once, using 3D Bresenham line algorithm l h Independent of vicinity radius Added irradiance based on distance between voxel location and occluding g voxel PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 7 Vicinity y Shading g Performance D ata D im s # D irections Tim e [m ins] Tim e/D ir.[s] Skull 256 x 256 x 203 312 14 3 14.3 28 2.8 Skull 256 x 256 x 203 1272 59.0 2.8 Cortex 128 x 512 x 256 1272 57.5 2.7 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 8 Dynamic Ambient Occlusion y Ropinski et al., Eurographics 2008 D iffuse Surface Shading D ynam ic A m bientO cclusion PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 9 Workflow 1 2 Capture Light Interactions Compress Lighting Information Preprocessing 3 Volume Rendering Interactive PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 10 Workflow 1 2 Capture Light Interactions Compress Lighting Information Preprocessing 3 Volume Rendering Interactive Light interactions … are captured for the vicinity only … are expressed by using a local histogram PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 11 Workflow 1 2 Capture Light Interactions Compress Lighting Information Preprocessing 3 Volume Rendering Interactive One local histogram for each voxel results i unmanageable bl d i in data sizes Light information is clustered to handle it interactively during rendering PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 12 Workflow 1 2 Capture Light Interactions Compress Lighting Information Preprocessing 3 Volume Rendering Interactive Rendering parameters are changed f l frequently Representative local histograms can be modulated interactively Volume rendering g requires q only y two additional texture fetches PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 13 Workflow 1 1 2 Capture Light Interactions Compress Lighting Information 3 Volume Rendering PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 14 2 3 Histogram Generation g 1 Analyze vicinity of each voxel Compute a normalized local histogram with bins PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 15 2 3 Histogram Generation g 1 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 16 2 3 Histogram Clustering g g Given: n local histograms (vectors) 1 Given: Desired:: m << n representatives (code Desired (code book)) book We exploit p a vector q quantization ((vq) q) for the clustering PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 17 2 3 Histogram Compression g p O Occurren nce Goal: generate a packed histogram 1 2 with i h j << i bins bi (i (i = initial i i i l number b off bins) bi ) Iterative splitting is used to reduce histogram dimensionality Example: p i = 12, 12, j = 9 Intensity PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 18 3 Interactive Rendering g 1 2 Two additional texture fetches required 1. Obtain the cluster ID of the current sample x 2. Fetch the current environment color is computed p by y considering g the current transfer function PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 19 3 Isosurface Shading g 1 Combination with the transfer function 0 intensity Apply Phong shading by using PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 20 2 3 2b Isosurface Shading g 1 Phong our technique nc=2048 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 21 2 3 Direct Volume Rendering g 1 Challenges 2 3 More than one hue is present Areas with participating media may occur Combination Co b a o with the e transfer a s e function u c o 0 intensity PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 22 2b Application Example pp p - DVR 1 nc=2048 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 23 2 3 Application Example pp p - DVR 1 Phong o g DAO O nc=2048 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 24 2 3 DAO Performance 1 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 25 2 3 Local Ambient Occlusion Hernell et al., Eurographics/IEEE VG 2007 D iffuse Surface Shading LocalA m bientO cclusion PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 26 Volumetric Ambient Occlusion Volumetric Object S Semi-transparent i t t ti tissues Global Light Source & Isotropic Incident Light Integrate Illumination for a point in the volume Accuracy: number of directions and resolution of light “cache” Volumetric Light Sphere reduce required directions PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 27 Additional Features Exploits Multiresolution LOD data Improves performance Reduces memory requirement No precomputation Interactive TF-based light emission Highlighting user-specific data ranges PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 28 LAO – single direction • Spherical neighbohod, Ω • Estimate attenuation of light • Semi-transparency PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 29 LAO – multiple directions • Compute LAO in K number of directions • Updated incrementally • one ray per frame PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 30 LAO Integration g Integration of irradiance from one direction Numerical i l evaluation l i Estimating light from K number of directions PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 31 Volumetric Light Cont ib tion Contribution Sof ftShadow ha Edges ge H ard Shadow Edges Boundary Only Contribution Volumetric Light Contribution PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 32 LAO with Emission X Affect intensity and color of x Emissive component in the TF Add CE (color light emission) to the integral PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 33 Multiresolution volume 162 Superfluous sampling in areas of low occlusion Multiresolution approach from Ljung et al., VolVis 2004 TF-based TF based LOD selection on flat blocking data structure LOD Selection 162 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 34 82 42 Packing g LOD Volume Packed texture High resolution (162) Middle resolution ((82) Low resolution (42) PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 35 Empty py Results Diffuse Illumination • NVIDIA GF8800 Ultra • 768 MB of graphics texture memory • 51 ms/frame to update one ray in the LAO map • 27 FPS LAO with 1 direction PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 36 LAO with 8 directions Results Emissive light in ray-casting Emissive light in LAO computations PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 37 Comparison w/ Diffuse Shading Diffuse illumination LAO, 32 rays PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 38 Application pp to Virtual Autopsy p y Case PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 39 Global Light Propagation g p g Hernell et al., Eurographics VG 2008 Approximate global illumination with first order scattering Extends LAO with initial light g p propagation p g step p PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 40 Pipeline: p Algorithm g Outline Change in Light Source / TF Intermediate results Local Piecewise Integration Global Shadow Volume αlp αg Global Light Contribution I + Scattering Volume ray casting Ieye (screen) PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 41 Piecewise Integration g αlp(x) ( ) α(n2) α(n ( 1) α(n0) PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 42 Piecewise Integration g αlp(s2) αlp(s1) αg(s) αlp(s0) PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 43 Improving p g Accuracy y of Direct Light g αg(s1) Id ((s)) αlp(s0) PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 44 Empty p y Space p Handling g PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 45 Speed-up global integration p p of g g Opacity is computed for each local segment originating at red points Gl b l integration Global i t ti using the precomputed piecewise integration can be b computed for fewer points Leading to a Shadow Volume Representation (SVR) Note that the integration for each point i still ill iis at hi high h PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION resolution 46 Shadow Volume Sensitivity y Original volume: 5123 voxels, Datareduction: 8.9:1, Segment Length: 16 voxels 323 643 1283 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 47 2563 Global+LAO Examples p PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 48 Compare p w/ / Diffuse Surface Gradient based local lighting Novel global + ambient lighting PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 49 Results: Translucency y Effects RΩ = 16, Ibias = 0 RΩ = 48, Ibias = 0 RΩ = 16, Ibias = 0.2 RΩ = 48, Ibias = 0.2 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 50 Performance Timing g Light Source / TF Change A Local Piecewise Integration milliseconds Global Shadow Volume B Global Light Contribution + Scattering S tt i milliseconds C Volume ray casting Original O i i l volume: l 5123 voxels l Datareduction: 8.9:1 Segment Length: 16 Viewport: 1024x1024 PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 51 Summary y Advanced interactive shading techniques Significantly improved realism Volumetric and iso-surface techniques q Complex structures and depth perception Color bleeding and emission Color-bleeding Global light propagation First order scattering …but maybe y we still need normals! PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 52 Final Views PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 53 References A. James Stewart. Vicinity Shading for Enhanced Perception of Volumetric Data. In Proc. IEEE Visualization 2003 Kevin M. Beason, Josh Grant, David C. Banks, Brad Futch, and M. Yousuff Hussaini. Pre-Computed Illumination for Isosurfaces. In Proc. of SPIE 2006 Ch i Wyman, Chris W Steven St Parker, P k Peter P t Shirley, Shi l and d Charles Ch l Hansen. H Interactive Display of Isosurfaces with Global Illumination. IEEE TVCG Vol. 12, No. 2, 2006 Timo Ropinski Ropinski, Jennis Meyer-Spradow, Meyer Spradow Stefan Diepenbrock Diepenbrock, Jörg Mensmann, and Klaus Hinrichs. Interactive Volume Rendering with Dynamic Ambient Occlusion and Color Bleeding. Computer p Forum (Eurographics ( g p 2008), ) Volume 27, Number 2, 2008 Graphics Eric Penner and Ross Mitchell. Isosurface Ambient Occlusion and Soft Shadows with Filterable Occlusion Maps. In Proc. of g p / Volume Graphics p 2008 Eurographics/IEEE PATRIC LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION 54 References 55 Frida Hernell, Patric Ljung, and Anders Ynnerman. Efficient Ambient and Emissive Tissue Illumination using g Local Occlusion in Multiresolution Volume Rendering. In Proc. of Eurographics/IEEE Volume Graphics 2007 Frida Hernell,, Patric Ljung, j g, and Anders Ynnerman. Interactive Global Light Propagation in Direct Volume Rendering using Local Piecewise Integration. In Proc. of Eurographics/IEEE Volume Graphics 2008 Patric Ljung, Claes Lundström, Ken Museth, and Anders Ynnerman. Transfer Function Based Adaptive Decompression for g of Large g Medical Data Sets. In Proc. of IEEE Volume Rendering Volume Visualization and Graphics 2004 Patric Ljung, Claes Lundström, and Anders Ynnerman. p in Direct Volume Multiresolution Interblock Interpolation Rendering. In Proc. of Eurographics/IEEE Symposium on Visualization 2006 Patric Ljung. Adaptive Sampling in Single Pass, GPU GPU-based based RaycastingPATRIC of Multiresolution Volumes. In Proc. LJUNG, SIEMENS CORPORATE RESEARCH, PRINCETON, USA of GPU-BASED VOLUME RAYCASTING WITH ADVANCED ILLUMINATION Eurographics/IEEE Volume Graphics 2006
