Motion Icon

1
Motion Icon
Feng Liu Computer Sciences Department University
of Wisconsin-Madison
2
Goal

• Motion Icon
• Summarize a motion
• capture data in a single image
• Application
• motion database browsing

From Assa 05
3
Solution

• Extract key frames
• Conveying as many motion info. as possible with
  few frames
• Render key frames
• Re-position frames
• Select proper camera settings to render them
  effectively

4
Key frame extraction

• Form posture clusters
• Select a representative frame from each cluster
  as a key frame

5
Re-position key frames

• Goal
• Avoid collision between each frames
• Solution
• Magnify the root trajectory by such a factor that
  there is no collision between consecutive frames
• Advantage
• Preserve the overall shape of motion trajectory

6
Proper camera setting selection

• Goal
• Render key frames in a way with minimal key frame
  occlusion
• Up vector 0 1 0
• At vector
• the center of the root trajectory
• Eye vector
• Eye-At line perpendicular to the plane determined
  by the the Up vector and the root trajectory

7
Camera settings
8
Demos

• Please refer to Webpage
• http//www.cs.wisc.edu/fliu/Course/cs777/p1/micon
  .htm

9
Conclusion

• A basic framework for creating motion icon
• Future work
• Use action lines and motion blur to convey motion
• Better pose clustering algorithm
• Extract key frames through cluster graph analysis
• Better human model than the stick model
• Dimension reduction