PatternBased Texturing Revisited Fabrice Neyret GRAVIRIMAGINRIA MariePaule Cani
Pattern-Based Texturing Revisited Fabrice Neyret / GRAVIR-IMAG-INRIA Marie-Paule Cani Grenoble, FRANCE
The texturing problem texturing without: • distortion • discontinuity • obvious periodicity
The texturing problem texturing without: • distortion • discontinuity • obvious periodicity
The texturing problem texturing without: • distortion • discontinuity • obvious periodicity
The texturing problem texturing without: • distortion • discontinuity • obvious periodicity
Texturing Methods • single image mapping [ Catmull 74 ] – need pre-distortion, large texture • pattern tiling – distortion, periodicity • interactive painting [ Haeberli+90 ] – large texture • procedural solid noise filtering – only for volumetric material surface idem, [ Perlin 85, Worley 96. . . ] + reaction-diffusion [Turk+91, Witkin+91] – distortion, large texture
Texturing Methods • single image mapping [ Catmull 74 ] – need pre-distortion, large texture • pattern tiling – distortion, periodicity • interactive painting [ Haeberli+90 ] – large texture • procedural solid noise filtering – only for volumetric material surface idem, [ Perlin 85, Worley 96. . . ] + reaction-diffusion [Turk+91, Witkin+91] – distortion, large texture
Texturing Methods • single image mapping [ Catmull 74 ] – need pre-distortion, large texture • pattern tiling – distortion, periodicity • interactive painting [ Haeberli+90 ] – large texture • procedural solid noise filtering – only for volumetric material surface idem, [ Perlin 85, Worley 96. . . ] + reaction-diffusion [Turk+91, Witkin+91] – distortion, large texture
Texturing Methods • single image mapping [ Catmull 74 ] – need pre-distortion, large texture • pattern tiling – distortion, periodicity • interactive painting [ Haeberli+90 ] – large texture • procedural solid noise filtering – only for volumetric material surface idem, [ Perlin 85, Worley 96. . . ] + reaction-diffusion [Turk+91, Witkin+91] – distortion, large texture
Texturing Methods • single image mapping [ Catmull 74 ] – need pre-distortion, large texture • pattern tiling – distortion, periodicity • interactive painting [ Haeberli+90 ] – large texture • procedural solid noise filtering – only for volumetric material surface idem, [ Perlin 85, Worley 96. . . ] + reaction-diffusion [Turk+91, Witkin+91] – distortion, large texture
Advantages of Pattern Tiling • reusability – material sample • high resolution with little memory • less designer work • standard – e. g. , graphics libraries
Drawbacks of Pattern Tiling • distortion [ LM 98. . . ] • discontinuity [ MYV 93 ] • obvious periodicity [ Stam 97 ]
Our Method • No global parameterization – tiling + continuity constraints • No distortion or discontinuity – equilateral triangular tiles • Texture tiling independent of geometry faces – texture mesh vs. geometry mesh
Our scheme geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Our scheme geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Our scheme geometric mesh build texture mesh project on surface [ Turk 92 ] parameterize patches match patterns textured mesh generate patterns
Our scheme geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Our scheme geometric mesh local u, v build texture mesh project on surface parameterize patches match patterns generate patterns local u, v textured mesh
Our scheme geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Our scheme geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Our scheme geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Our scheme geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Local Parameterization geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Local Parameterization • geodesic distance to border (shortest path on the surface) – flood fill from each border • normalize • Levy and Mallet [98] • Eck et al. [95] (harmonic maps) • Lee et al. [98] (MAPS) d 2 Other choices d 1 d 3 Geodesic barycentric coordinates
Matching of Patterns geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Matching of Patterns • • Escher (1898 -1972) Penrose (1931 -. . . ) • Graph coloring • Stam [97] M. C. Escher 1943
Matching of Patterns vertex Boundary Condition edge Boundary Condition central zone
Matching of Patterns vertex Boundary Condition edge Boundary central Condition zone
Matching of Patterns vertex Boundary Condition edge Boundary central Condition zone
Matching of Patterns vertex Boundary Condition edge Boundary central Condition zone Pattern set
Matching of Patterns vertex Boundary Condition edge Boundary central Condition zone
Tunings yielding 4 patterns 1 vertex Boundary Condition 1 edge BC per pair of vertex 2 constraints E and E per edge 2 kinds of edge constraint 4 possible triples of edges 1 central zone per triple of edges 4 patterns
Tunings yielding 4 patterns 1 vertex Boundary Condition 1 edge BC per pair of vertex 2 constraints E and E per edge 2 kinds of edge constraint 4 possible triples of edges 1 central zone per triple of edges 4 patterns
Tunings yielding 4 patterns 1 vertex Boundary Condition E 1 edge BC per pair of vertex 2 kinds of edge constraint E E E E 4 patterns E 4 possible triples of edges 1 central zone per triple of edges E E 2 constraints E and E per edge E
Tunings yielding 4 patterns 1 vertex Boundary Condition E 1 edge BC per pair of vertex 2 kinds of edge constraint E E E E 4 patterns E 4 possible triples of edges 1 central zone per triple of edges E E 2 constraints E and E per edge E
Tunings yielding 4 patterns 1 vertex Boundary Condition 1 edge BC per pair of vertex 1 central zone per triple of edges
Tunings yielding 4 patterns 1 vertex Boundary Condition 1 sym edge BC per pair of vertex 1 kind of edge constraint (E = E) 1 triple of edges 4 central zones per triple of edges
Tunings yielding 4 patterns 1 vertex Boundary Condition 2 sym edges BC per pair of vertex 4 possible triples of edges 1 central zone per triple of edges
Tunings yielding 4 patterns 2 vertex Boundary Conditions 1 sym edge BC per pair of vertex 3 kinds of edge constraint 4 possible triples of compatible edges 1 central zone per triple of edges
Pattern Generation geometric mesh build texture mesh project on surface parameterize patches match patterns textured mesh generate patterns
Pattern Generation EE • drawing EE EE • photo editing EE EE Images:
Pattern Generation Images: • photo editing • drawing
Pattern Generation Procedural: • Perlin • Worley
Video now, the video
Conclusions A new texturing approach • with little distortion, little discontinuity, no periodicity • pattern-based, thus less work, high resolution, reusable samples • compatible with real-time rendering
Future Work • multiscale patterns – handles and legs • heterogeneity – varying pattern occurrence – varying tile size (field, curvature) • anisotropy – tensor field – open problem!
- Slides: 46