1. Mesh Modelling With Curve Analogies
Steve Zelinka Michael Garland
University of Illinois at Urbana-Champaign

2. In a Nutshell

3. Overview Motivation Related Work Details of our Approach Results
Curve Selection
Surface Transformation
Results
Future Work

4. Motivation Reduce artistic skill required for modelling
Solution: Modelling by Analogy A : A’ :: B : ?
Images [Hertzmann et al 2001]
Curves [Hertzmann et al 2002]

5. Mesh Analogies?
User burden
Unsolved technical issues
: :: : ?

6. Related Work Wires [Singh and Fiume 1998]
Excellent control over editing
Highly skilled artist required

7. Related Work Teddy [Igarashi et al 1999]
Intuitive sketch-based interface
Limited class of models

8. Approach Overview Select surface curves
Transform surface curves with Curve Analogies
Transform the surface
2D sketch-based manipulation
Simple implementation

9. Curve Selection Planar intersection curves Parallel or rotating slices
Orthogonal to skeleton

10. Curve Selection Planar intersection curves Silhouette curves
Parallel or rotating slices
Orthogonal to skeleton
Silhouette curves

11. Generality Issues Features controlled only on and along curves
Use orthogonal, intersecting sets of curves
Multiple passes

12. Curve Analogies User sketches unfiltered, filtered curves
Identical parameterizations required
Based on joint neighbourhood matching
Neighbourhoods must be aligned before comparison

13. Surface Transformation
Similar to Wires
Vertices near a curve track movement of their closest points on the curve

14. Surface Transformation
Similar to Wires
Vertices near a curve track movement of their closest points on the curve
Vertex movement inversely proportional to distance to curve

15. Influence Radius
Radius of influence of each curve can be varied

16. Influence Radius
Radius of influence of each curve can be varied

17. Influence Radius
Radius of influence of each curve can be varied

18. Multiple Curves Vertices can be influenced by multiple curves
Candidate position from each influencing curve
Final position weighted average of candidates

19. Results Curve Analogies Dominate compute time
Can be difficult to control

20. Future Directions Better Curve Analogies
Avoid orientation flipping using surface information
Use intrinsic curve parameterization to accelerate

21. Future Directions More curve families Iso-parameter curves
Signal-specific curves
Suggestive contours [DeCarlo et al 2003]

22. Future Directions Influence radius
Use spatially-based multi-analogies
Adaptive setting
Dynamics-based Surface Transformation
Prevent self-intersections
Allow topology changes

23. Thanks Funded in part by a grant from the NSF (CCR-0086084)
Contact: Steve Zelinka
Michael Garland