Image Registration Method BSpline 3¶
If you are not familiar with the SimpleITK registration framework we recommend that you read the registration overview before continuing with the example.
This example performs registration of multi-modality images with a multi-resolution BSpline approach.
A BSplineTransform usually has a large number of parameters which increases the complexity and duration of optimizing the deformation. The multi-resolution BSpline approach initially performs the registration at a lower resolution with fewer parameters at the first level and then adapts or resamples the BSpline control points to a higher resolution at subsequent levels. This transformation adaption is done concurrently with the ImageRegistrationMethod’s multi-level feature. This enables the setting of the shrink factor, smoothing sigmas, sampling percentage and BSpline resolution to vary per level to efficiently solve a diverse set of registration problems.
The initial transform is the low resolution BSpline. The scaling
factor of each BSpline used per level is specified with the
ImageRegistration::SetInitialBSpline method’s third argument as an
integer array. The first value is usually 1 and it is reasonable to
double the resolution at each registration level. For this example the
last resolution is 5, so that the result is comparable to the
transformation from the previous example.
It can be important to monitor and observe the transform at each level or iteration. When the “inplace” option is enabled, the transform passed as the initial transform will be up to date during the registration process which enables it to be used in event commands.