1. Computer Generated Watercolor
Curtis, Anderson, Seims, Fleisher, Salesin
SIGGRAPH 1997
Presented by
Yann SEMET
Universite of Illinois at Urbana Champaign
Universite de Technologie de Compiegne

2. Background
 NPR
 Purpose : aesthetic rather than
technical
 Artificial art ?

3. Harold Cohen – 80’s

4. Haeberli - 1990

5. Meier - 1995

6. Litwinowicz - 1997

7. Hertzmann – 1998, 2001

8. Gooch - 2001

9. Today : Curtis et al. - 1997

10. Overview
 Particularities of Watercolor
 Computer simulation
 Fluid simulation
 Kubelka-Munk rendering
 Applications
 Discussion

11. Like no other medium
 Beautiful textures and patterns
 Reveals the motion of water
 Luminous, glowing

12. Blake (1757-1827)

13. Turner (1775-1851)

14. Constable (1776-1837)

15. Cezanne (1839-1906)

16. Kandinski (1866-1944)

17. Klee (1879-1940)

18. Carter (1955-)

19. Watercolor materials
 Paper
 Pigments

20. Watercolor effects
a) Dry brush d) Granulation
b) Edge darkening e) Flow
c) Back runs f) Glazing

21. Simulation..

22. Fluid simulation I
 3 layers :

23. Fluid simulation II
 Parameters of the simulation :
 Wet-area mask : M
 Velocities : u,v
 Pressure : p
 Concentration : gk
 Height of paper : h
 Physical properties : density, staining power,
granularity, etc.
 Fluid properties : saturation, capacity, etc.

24. Paper simulation
 Supposedly : shape of every fiber
matters
 A simpler model : a height field
 Generation : Perlin’s noise and Worley’s
cellular textures

25. Main loop
 For each time step
 Move Water
 Update velocities
 Relax Divergence
 Flow Outward
 Move Pigment
 Transfer Pigment
 Simulate Capillary Flow

26. Conditions for realism
 Flow must be constrained so water
remains within M
 Surplus of water causes flow outward
 Flow must be damped to minimize
oscillating waves
 Flow is perturbed by texture of paper
 Local changes have global effects
 Outward flow to darken edges

27. Rendering : Kubelka-Munk
 For each pigment, 2 coeff. Per RGB layer :
 K : absorbtion
 S : scattering
 Supposedly : K and S are measured
 Here : user provides Rw and Rb

28. Types of paints
 Opaque (e.g. Indian Red)
 Transparent (e.g. Quinacridone Rose)
 Interference (e.g. Interference Lilac)
 Different hues (e.g. Hansa Yellow)

29. Optical compositing
 Compute R and T :
 Then compose :
 Weight relatively to relative thicknesses

30. Discussion of the KM model
 Assumptions partially satisfied :
 Identical refractive indices
 Random orientation of pigments
 Diffuse illumination
 1 wavelength at a time
 No chemical interaction
 Works surprisingly well !
 OK, because we’re looking for appearance,
not actual modeling

31. Application I
 Interactive painting :

32. Application II
 Watercolorization :

33. Application III
 3D models :

34. Future work
 Other effects
 Automatic rendering
 Generalization
 Animation

35. Summary
 A particular painting technique
 A physically based simulation
 Fluid motion
 Optical compositing
 Application and results

36. Conclusion and discussion
 Efficiency issues and long term interest
 Border between art, physics and
computer science