extracting curve- skeletons from digital shapes using occluding contours
DESCRIPTION
Extracting Curve- Skeletons from Digital Shapes Using Occluding Contours. M. Livesu and R . Scateni. Università degli Studi di Cagliari ( Italy ). Why ?. Graph-like structure. 1D!. High- level representation. [Au et al, 2010]. [Jiang et al, 2012]. Computation. ✓. 1D - PowerPoint PPT PresentationTRANSCRIPT
Extracting Curve-Skeletons from Digital Shapes Using Occluding Contours
M. Livesu and R. Scateni
Università degli Studi di Cagliari (Italy)
Why?
1D!
[Jiang et al, 2012]
High-level representation [Au et al, 2010]
Graph-like structure
Computation
• 1D• Graph-like structure
✗
✓
• 1D• Graph-like structure
State of the art (input-wise)
[Cornea et al, 2005]
Voxels
[Liu et al, 2010]
[Dey and Sun, 2006]
[Au et al, 2008]
[Tagliasacchi et al, 2012]
Triangle meshes Polygon meshesPolygon soups
Implicit surfaces….
Point clouds
[Tagliasacchi et al, 2009]
[Cao et al, 2010]
[Sharf et al, 2007]
Sometimes resolution is more important than theshape itself…
[Au et al, 2008] [Dey and Sun, 2006]
Coarse hand (≈1k triangles)
State of the art (resolution-wise)
Perceptual approach
What are the alternatives? Focus on the appearance of the shapes
1K triangles
273K triangles
GOALS: resolution and format independence!
Occlusions
MINOCCLUSION
MAXOCCLUSION
Pipeline
gather silhouettes
filter 2D skeletons
back-project medial points
Compute skeleton paths
Skeleton filtering
Skeleton filtering
Skeleton filtering
Skeleton filtering
✗
✓
Skeleton filtering
Skeleton filtering
Skeleton filtering
MINOCCLUSION
MAXOCCLUSION
Back-projection
Back-projection
Curve Reconstruction
1D Moving Least Squares [Lee, 2000]
Conclusions
• Silhouette space
• Format independence
• Resolution independence
Questions