INSTITUTE OF COMPUTING TECHNOLOGYCHINESE ACADEMY OF SCIENCES 1 A Pose-Independent Method of Animating Scanned Human Bodies Yong Yu, Tianlu Mao, Shihong Xia, Zhaoqi Wang Institute of Computing Technology, Chinese Academy of Sciences
INSTITUTE OF COMPUTING TECHNOLOGY 2 Outlines Goal Previous Methods Our Method Segmentation Skeleton Extraction Skin Deformation Results Conclusion & Future Work
INSTITUTE OF COMPUTING TECHNOLOGY 3 Goal To animate scanned human bodies
INSTITUTE OF COMPUTING TECHNOLOGY 4 Previous Methods Model Mapping [ Seo03 ] To map a skeleton and skin deformation of a template model to a scanned human body Pose-independent, require a template model
INSTITUTE OF COMPUTING TECHNOLOGY 5 Previous Methods Model Intersecting [Ju00, João03] To intersect scanned body into contours with horizontal planes. To extract joints according to perimeters or average radiuses of the contours. Automatic, pose-dependent
INSTITUTE OF COMPUTING TECHNOLOGY 6 Previous Methods Model Segmentation [Xiao03, Werghi06] To utilize Morse theory to segment scanned human body Pose-independent, only segmentation
INSTITUTE OF COMPUTING TECHNOLOGY 7 Previous Methods Puppet Rigging [Baran07] To rig and animate 3D puppets Pose-independent, orientation-dependent
INSTITUTE OF COMPUTING TECHNOLOGY 8 Key Issues Pose-independent How to animate scanned human body in different poses or in different orientations? Automatic Does the method perform automatically?
INSTITUTE OF COMPUTING TECHNOLOGY 9 Animating scanned body Segmentation Skeleton Extraction Skin Deformation
INSTITUTE OF COMPUTING TECHNOLOGY 10 Geodesic distance from x to a source point Morse Functions Height function [Ju 00] Sum of geodesic distances from x to all points [Werghi 06]
INSTITUTE OF COMPUTING TECHNOLOGY 11 Segmentation Source Point Candidate Considering symmetry Bilateral symmetry Front and back symmetry Both symmetry
INSTITUTE OF COMPUTING TECHNOLOGY 12 Segmentation - source point Source point: Top of head Source point: Crotch vs
INSTITUTE OF COMPUTING TECHNOLOGY 13 Segmentation Source Point and Feature Points ab c ef g Source Point
INSTITUTE OF COMPUTING TECHNOLOGY 14 Segmentation Morse function isolines, Contours Topological structureSegments
INSTITUTE OF COMPUTING TECHNOLOGY 15 Skeleton Extraction Assumption : Contours at joints are irregular ; the other contours are more regular and more similar to a circle Circularity Function (to estimate similarity degree between a contour and a circle) Joint Extraction M joint is a joint contour Circularity function gains local minimum in joint contour whose center is a true joint
INSTITUTE OF COMPUTING TECHNOLOGY 16 Skeleton Extraction
INSTITUTE OF COMPUTING TECHNOLOGY 17 Skin Deformation Skeleton Subspace Deformation Separating Contours Joint Contours – for all joints except shoulders and thighs Additional Contours – for shoulders and thighs
INSTITUTE OF COMPUTING TECHNOLOGY 18 Joint Contours M J : Joint Contour M 1 :Region Border M 2 : Region Border Deformation Region Skin deformation weights
INSTITUTE OF COMPUTING TECHNOLOGY 19 Additional Contours a b cd e SegmentsErrors in Segmentation
INSTITUTE OF COMPUTING TECHNOLOGY 20 Results - Various Human Bodies
INSTITUTE OF COMPUTING TECHNOLOGY 21 Results - Various Postures
INSTITUTE OF COMPUTING TECHNOLOGY 22 Results - Various Orientations
INSTITUTE OF COMPUTING TECHNOLOGY 23 Results Time 5000 Faces Faces Faces Faces Source Point Extraction0.5s1.4s5.1s14.2s Segmentation0.1s0.3s0.9s1.4s Joint Extraction and Skin Deformation 0.3s0.8s4.4s8.8s Total0.9s2.5s10.4s24.4s
INSTITUTE OF COMPUTING TECHNOLOGY 24 Conclusion and Future Work Conclusion Pose-independent Automatic Robust Future Work Loop in topological structure of body
INSTITUTE OF COMPUTING TECHNOLOGYCHINESE ACADEMY OF SCIENCES 25 Thank you