14.08.2013 Views

Docteur de l'université Automatic Segmentation and Shape Analysis ...

Docteur de l'université Automatic Segmentation and Shape Analysis ...

Docteur de l'université Automatic Segmentation and Shape Analysis ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Chapter 4 Statistical shape mo<strong>de</strong>l of Hippocampus 79<br />

where dΩ = sin θdθdφ is the surface element on S 2 . The condition of isometry is<br />

satisfied for rotation Γ<br />

fi ◦ Γ − fj ◦ Γ 2 ∫<br />

=<br />

∫<br />

=<br />

∫<br />

=<br />

S 2<br />

S 2<br />

S 2<br />

fi(Γ(θ, φ)) − fj(Γ(θ, φ)) 2 dΩ<br />

fi(Γ(θ, φ)) − fj(Γ(θ, φ)) 2 J −1 (Γ(θ, φ))dΩ(Γ(θ, φ))<br />

fi(Γ(θ, φ)) − fj(Γ(θ, φ)) 2 dΩ(Γ(θ, φ))<br />

= fi − fj 2<br />

(4.17)<br />

since the Jacobian J for rotation is 1 on S 2 . Assuming the shape surfaces {Xi} are<br />

aligned (e.g. by ICP algorithm) via similarity transforms to a common template,<br />

the rotation Γi can be found by minimizing the pairwise distance<br />

Γi = arg min fi ◦ Γ − f1<br />

Γ∈SO(3)<br />

2 , (4.18)<br />

where f1 is an arbitrarily selected parameterization as the template. Multiple runs<br />

with different choices of template may be performed in or<strong>de</strong>r to avoid bias towards<br />

the choice of template.<br />

In implementation, the L2 distance may be evaluated by sampling on S 2<br />

fi − fj 2 4π<br />

k<br />

k∑<br />

fi(θs, φs) − fj(θs, φs)<br />

s=1<br />

2 , (4.19)<br />

where (θs, φs), s = 1, · · · k are uniformly sampled on S 2 by Platonic solids such<br />

as octahedron, do<strong>de</strong>cahedron or icosahedron. By parameterizing the rotation Γ<br />

using Rodrigues parameters or quaternions, the optimization problem in (4.18)<br />

thus becomes a least-square problem with closed form solution, which can be<br />

solved by Nel<strong>de</strong>r-Mead simplex or other st<strong>and</strong>ard optimizers.<br />

A multi-resolution scheme can be easily implemented by subdividing the initial<br />

sampling (Figure 4.3). The multi-resolution algorithm for rotational reparameter-<br />

ization is listed in Algorithm 7.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!