Equipment Quickened Versatile EWA Volume Splatting.


91 views
Uploaded on:
Description
Equipment Quickened Versatile EWA Volume Splatting. Wei Chen ZJU Liu Ren CMU Matthias Zwicker MIT Hanspeter Pfister MERL. Volume Splatting. Object-request technique 3D reproduction part focused at each voxel (circular Gaussian)
Transcripts
Slide 1

Equipment Accelerated Adaptive EWA Volume Splatting Wei Chen ZJU Liu Ren CMU Matthias Zwicker MIT Hanspeter Pfister MERL

Slide 2

Volume Splatting Object-request system 3D recreation piece focused at each voxel (curved Gaussian) Voxel commitment = 2D foot shaped impression (shading, darkness) Weighted foot shaped impressions amassed into picture 2D foot shaped impressions = splats Screen Voxel parts

Slide 3

Related Work Quality Westover1989 Crawfis 1993 EWA Swan 1997 Mueller 1999 Image-adjusted Huang 2000 Zwicker 2001 Our work Swan Xue 2003 Axis-adjusted Speed Fast splats Texture splats OpenGL ex Software

Slide 4

Outline EWA volume splatting Adaptive EWA splatting GPU usage Results and conclusions

Slide 5

EWA Volume Splatting Low-Pass Filter Projection W Convolution EWA volume resampling channel Compensate associating ancient rarities because of point of view projection EWA Filter = low-pass channel twisted remaking channel Volume

Slide 6

EWA Volume Splatting (512x512x3) Reconstruction channel just: 6.25 fps EWA channel: 4.97 fps Low-pass channel just: 6.14 fps EWA channel: 3.79 fps

Slide 7

Analysis of EWA Filter Minification Warped recon-struction bit Low-pass channel Resampling channel Magnification

Slide 8

Analysis of EWA Filter Shape of EWA Splat is reliant on separation from the perspective plane r h Low-pass channel range r k Reconstruction channel sweep EWA splat u 2 Distance to the perspective plane Note that

Slide 9

Adaptive EWA Filtering Warped recon-struction portion Low-pass channel Resampling channel if u 2 > An utilization low-pass channel if A<u 2 < B use EWA channel if u 2 < B use reproduction channel

Slide 10

Patch Processing Process a 8 x 8 patch of voxels at once Filter determination in light of four corners of every patch (pick littlest) Traversal request Patch Distance

Slide 11

Adaptive EWA Volume Splatting (512x512x3) Adaptive EWA channel: 6.88 fps EWA channel: 4.83 fps Adaptive EWA channel: 1.84 fps EWA channel: 1.75 fps

Slide 12

Outline EWA volume splatting Adaptive EWA splatting GPU execution Results and conclusions

Slide 13

Object-Space EWA Splatting Projection Vertex shader calculation Texture mapping Textured quad Object-space EWA splatting with composition mapping [Ren et al. Eurographics 2002] EWA Splat (circular Gaussian) Texture (unit Gaussian) (1,1) (0,1) (0,0) (1,0) Unit quad

Slide 14

Proxy Geometry Template . . . Rectilinear volumes: utilize one intermediary geometry layout for all cuts in every bearing Store vertex files in AGP memory Regularity Voxel geometry Quad geometry Proxy geometry format

Slide 15

Vertex Compression Compress every vertex to 32 bits Decompression on-the-fly in programmable equipment To store vertex data of 256x256x256 volume in feature memory Without pressure 2,048 MBytes  With pressure 12 MBytes  Retained-mode equipment speeding up practical

Slide 16

Retained versus Quick Mode Factor of ~10 change

Slide 17

Interactive Classification: Opacity Culling Hardware-quickened rundown based traversal For every cut For each 32 x 32 patch of voxels (littler lists) Indices of intermediary geometry composed into iso-quality records utilizing pail sort; CPU consolidations records on-line Render just iso-worth records with noticeable voxels 0 128 Patch 256

Slide 18

Interactive Classification: Opacity Culling Includes changes to TF each edge Factor of ~10 change

Slide 19

Deferred Shading Volume surface access is just conceivable in section programs * However, per piece shading is lavish Solution: conceded shading in two passes * Newer GPUs permit composition access in vertex programs

Slide 20

Deferred Shading Pass one: 3D surface access, characterization and enlightenment in vertex shader, render one pixel for every voxel Pass two: reuse the pixel information from the first go to shade the 2D foot shaped impression Performance pick up: 5%-10% speedup Pass one Pass two Final result

Slide 21

Experiments P4 2.4 GHz ATI 9800 Pro with 256 MB RAM Direct3D 9.0b with VS 2.0 and PS 2.0 Vertex shader guidelines

Slide 22

Sheet-cushion Composition 0.80 fps 3.00 fps 3.45 fps Axis-adjusted traversal, expansion in sheet cradles, then mixing front-to-back

Slide 23

UNC Head: 208x256x225 #Rendered splats: 2,955,242 2.86 fps 8.5M splats/sec

Slide 24

Bonsai: 256x256x128 #Rendered splats: 274,866 7.53 fps 2M splats/sec

Slide 25

Engine: 256x256x110 #Rendered splats: 247,577 10.28 fps 2.5M splats/sec

Slide 26

Lobster: 301x324x56 #Rendered splats: 555,976 10.60 fps 5.9M splats/sec

Slide 27

Video

Slide 28

Our Contributions Adaptive EWA calculation Volume information pressure Retained-mode equipment increasing speed Interactive darkness separating Deferred two-pass shading

Slide 29

Future Work Image-adjusted EWA volume splatting Irregular volume splatting Pointsprites in OpenGL Floating point surfaces Vertex composition for grouping

Slide 30

Acknowledgments Jessica Hodgins (CMU) Markus Gross (ETH) http://graphics.cs.cmu.edu/ventures/adpewa/index.html

Slide 31

.:tslid

Recommended
View more...