Biomedical Engineering Reference
In-Depth Information
a
= 5000
a
= 1000
a
= 500
= 250
= 100
= 50
a
a
a
Figure 8.11: Results of the registration without robust estimator. The differ-
ent volumes correspond to different values of the parameter
α
, and must be
compared to the source volume.
We then perform the robust multigrid registration process, with a robust
function only on the regularization term. Results are presented in Fig. 8.12, with
two “extreme” values of the parameter
α
. In that case, the modification of the
topology is possible, while preserving the global smoothness of the solution.
However, the aperture problem is still present in the tubular structure on the
right. This experiment makes it possible to verify the link between the introduc-
tion of a robust estimator on the regularization term and the possibility to handle
local change of topology. In addition, the robust registration process appears to
be also more robust with respect to the parameter
α
, because the results of the
registration are very similar, when
α
varies in a range of [100, 3,000].
8.3.3.3
Experiment on Two Subjects
Importance of Intensity Correction.
We first want to present cases
where the registration method cannot work properly without a prior intensity
