Geometry Images Steven Gortler Harvard University Steven Gortler Harvard University Xianfeng Gu Harvard University Xianfeng Gu Harvard University Hugues Hoppe Microsoft Research Hugues Hoppe Microsoft Research
Irregular meshes Vertex 1 x 1 y 1 z 1 Vertex 2 x 2 y 2 z 2 … Face Face …
Texture mapping Vertex 1 x 1 y 1 z 1 Vertex 2 x 2 y 2 z 2 … s1 t1s1 t1s2 t2s2 t2s1 t1s1 t1s2 t2s2 t2 normal map s t Face Face …
Complicated rendering process Vertex 1 x 1 y 1 z 1 Vertex 2 x 2 y 2 z 2 … random access! s1 t1s1 t1s2 t2s2 t2s1 t1s1 t1s2 t2s2 t2 Face Face … ~40M Δ/sec
Semi-regular representations irregular vertex indices only semi-regular [Eck et al 1995] [Lee et al 1998] [Khodakovsky 2000] [Guskov et al 2000] …
Geometry Image geometry image 257 x 257; 12 bits/channel 3D geometry completely regular sampling
Basic idea demo cut parametrize
Basic idea cut sample
cut [r,g,b] = [x,y,z] render store
How to cut ? sphere in 3D 2D surface disk
How to cut ? l Genus-0 surface any tree of edges sphere in 3D 2D surface disk
How to cut ? l Genus-g surface 2g generator loops minimum torus (genus 1)
Surface cutting algorithm (1) Find topologically-sufficient cut: 2g loops [Dey and Schipper 1995] [Erickson and Har-Peled 2002] (2) Allow better parametrization: additional cut paths [Sheffer 2002] (1) Find topologically-sufficient cut: 2g loops [Dey and Schipper 1995] [Erickson and Har-Peled 2002] (2) Allow better parametrization: additional cut paths [Sheffer 2002]
Step 1: Find topologically-sufficient cut (a) retract 2-simplices (b) retract 1-simplices
Results of Step 1 genus 6 genus 0 genus 3
Step 2: Augment cut l Make the cut pass through “extrema” (note: not local phenomena). l Approach: parametrize and look for “bad” areas. l Make the cut pass through “extrema” (note: not local phenomena). l Approach: parametrize and look for “bad” areas.
Step 2: Augment cut …iterate while parametrization improves
Results of Steps 1 & 2 genus 1 genus 0
Parametrize boundary Constraints: n cut-path mates identical length n endpoints at grid points Constraints: n cut-path mates identical length n endpoints at grid points a a’ a a’ no cracks
Parametrize interior –optimizes point-sampled approx. [Sander et al 2002] n Geometric-stretch metric –minimizes undersampling [Sander et al 2001] n Geometric-stretch metric –minimizes undersampling [Sander et al 2001]
Previous metrics (Floater, harmonic, uniform, …) Stretch parametrization
SampleSample geometry image
RenderingRendering (65x65 geometry image)
rendering geometry image x 12b/ch normal-map image x 8b/ch Rendering with attributes
Advantages for hardware rendering l Regular sampling no vertex indices. l Unified parametrization no texture coordinates. Raster-scan traversal of source data: geometry & attribute samples in lockstep. Summary: compact, regular, no indirection Summary: compact, regular, no indirection l Regular sampling no vertex indices. l Unified parametrization no texture coordinates. Raster-scan traversal of source data: geometry & attribute samples in lockstep. Summary: compact, regular, no indirection Summary: compact, regular, no indirection
normal map 512x512; 8b/ch Normal-Mapped Demo geometry image 129x129; 12b/ch demo
color map 512x512; 8b/ch Pre-shaded Demo geometry image 129x129; 12b/ch
ResultsResults 257x257 normal-map 512x512
ResultsResults 257x257 color image 512x512
Mip-mappingMip-mapping 257x257129x12965x65 boundary constraints set for size 65x65
Hierarchical culling view-frustum culling backface culling geometry image normal-map image
CompressionCompression 1.5 KB + topological sideband (12 B) fused cut 295 KB Image wavelet-coder
Compression results 1.5 KB 3 KB 12 KB 49 KB 295 KB
Rate distortion
Some artifacts aliasing anisotropic sampling
SummarySummary l Simple rendering: compact, no indirection, raster-scan stream. l Mipmapped geometry l Hierarchical culling l Compressible l Simple rendering: compact, no indirection, raster-scan stream. l Mipmapped geometry l Hierarchical culling l Compressible
Future work l Better cutting algorithms l Feature-sensitive remeshing l Tangent-frame compression l Bilinear and bicubic rendering l Build hardware l Better cutting algorithms l Feature-sensitive remeshing l Tangent-frame compression l Bilinear and bicubic rendering l Build hardware