Posing Guide: 342 female poses, 61 male poses, 46 pair poses
Rigid Registration of Different Poses of Animated Shapes
description
Transcript of Rigid Registration of Different Poses of Animated Shapes
![Page 1: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/1.jpg)
Rigid Registration of Different Poses of Animated Shapes
M. Livesu and R. Scateni
University of Cagliari
![Page 2: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/2.jpg)
[Thiery et al, 2013]
[De Aguiar et al, 2008][Sumner and Popovic, 2004]
[Marras et al, 2012]
Motivation
[Winkler et al, 2010]
![Page 3: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/3.jpg)
Problem statement
is an isometry
![Page 4: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/4.jpg)
Challenges
Poses do not match perfectly• Focus on rigid areas• Focus on a compact area
![Page 5: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/5.jpg)
State of the art
PCA based algorithms
[Chaouch et al, 2008]
[Aiger et al, 2008]
ICP based algorithms
Randomized algorithms
Global!
Require a good start solution!
Global!
Requires high overlapping (> 40%)!
![Page 6: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/6.jpg)
The algorithm
Select a Region of Interest (ROI)
Register poses by aligning ROIs
![Page 7: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/7.jpg)
ROI detection
0
MAX• Detect seeds ( )
• Max surface
• Region growing
[Meyer et al, 2002]
![Page 8: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/8.jpg)
Registration
[Sorkine, 2007]
![Page 9: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/9.jpg)
Results
26,907 TOT vertices1,092 ROI vertices174ms
7,207 TOT vertices965 ROI vertices59ms
7,207 TOT vertices544 ROI vertices56ms
![Page 10: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/10.jpg)
Multiple poses
• All vs All
• Chain
• All vs 1
![Page 11: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/11.jpg)
Conclusions
• Based on Gaussian Curvature
• Easy to implement
• Multiple poses
https://sites.google.com/site/marcolivesu/Download the source code at:
![Page 12: Rigid Registration of Different Poses of Animated Shapes](https://reader035.fdocuments.net/reader035/viewer/2022062500/56815ab0550346895dc85992/html5/thumbnails/12.jpg)
Questions