Automatic Registration for Articulated Shapes |
|||
University of California, San Diego |
|||
![]() |
|||
Registration for an arm dataset pair. The source mesh (a) is aligned to
the target mesh (b). The hand region is missing a significant amount of
data in both meshes, but after alignment the surface of the hand is
completed nicely (c). The assigned labels are shown in (d) for the
source (bottom) and the target (top), and corresponding parts have the
same label assignment. Also, the segmentation naturally corresponds to
the different parts of the arm.
|
|||
AbstractWe present an unsupervised algorithm for aligning a pair of shapes in the presence of significant articulated motion and missing data, while assuming no knowledge of a template, user-placed markers, segmentation, or the skeletal structure of the shape. We explicitly sample the motion, which gives a priori the set of possible rigid transformations between parts of the shapes. This transforms the problem into a discrete labeling problem, where the goal is to find an optimal assignment of transformations for aligning the shapes. We then apply graph cuts to optimize a novel cost function, which encodes a preference for a consistent motion assignment from both source to target and target to source. We demonstrate the robustness of our method by aligning several synthetic and real-world datasets. DownloadsPaper
[ Paper ] (16.3 MB Adobe PDF, Hi-Res)
Talk Slides[ Slides ] (15.7 MB Powerpoint Presentation) |