Image Processing Reference
Figure 6.4: Comparison of shape and trajectory spaces. (a) In traditional approaches, a 3D object is
represented as a point in shape space. (b) By contrast, in the approach of Akhter et al. [ 2008 ], the trajectory
is represented by a single point in trajectory space. Courtesy of I. Akhter.
Torresani et al.
Xiao et al.
Figure 6.5: Comparison of several approaches on a dance sequence from the CMU mocap database.
The images were based on Akhter et al. [ 2008 ]. Note that their method produces more accurate results
than the others. Black dots represent the ground-truth points, whereas gray circles are the reconstructed
ones. Courtesy of I. Akhter
In addition to its originality, this method has the advantage of only requiring orthonormality
constraints as additional knowledge to yield accurate reconstruction. This is due to the fact that the
basis is fixed, and therefore fewer unknowns need be determined in the reconstruction process.