[Insight-users] problem with registration of binay images

[Ming Chao]

Hi,

I encountered a problem when doing registration between two binary images.
The organ is black and the background is white. The 2nd image is just a
translational transform of the first image.  Along the edge of the organ I
selected a bunch of regions with the centers on the edge. The size of the
region was tested with different values. The following are some of the
outputs with certain registration configuration.

(1)
  typedef itk::BSplineInterpolateImageFunction<ImageType, double >
InterpolatorType;
  typedef itk::ImageRegistrationMethod<ImageType, ImageType >
RegistrationType;
  typedef itk::NormalizedCorrelationImageToImageMetric<ImageType,ImageType
>    MetricType;
  typedef itk::OnePlusOneEvolutionaryOptimizer       OptimizerType;
  typedef itk::TranslationTransform< double, Dimension >     TransformType;

The output of the iteration for one region looks like:
# of iterations   metric          position
1                       -0.875595   [-85.8273, -54.5089]
2                       -0.875595   [-85.8273, -54.5089]
3                       -0.875595   [-81.0874, -50.372]
4                       -0.875595   [- 81.0874, -50.372]
...
17                     -0.875595   [-81.0874, -50.372]
18                     -0.875595   [-82.8751, -49.8843]
19                     -0.875595   [-82.8751, -49.8843]
...
284                  -0.875595   [-84.5175, -63.2061]
285                  -0.875595   [-84.8451, -63.7505]
286                  -0.875595   [-84.8451, -63.7505]
...
506                  -0.875595   [-84.4683, -63.7624]
507                  -0.875595   [-84.4683, -63.7624]

After 507 iterations, the postion changes but the metric remains the same.
All other regions are pretty much the same.


(2) . Same as (1) except the optimizer is changed to LBFGSBOptimizer. The
typical observation is that for almost each region there is no iteration at
all. The position parameters are just the initial values.

(3) Same as (2) and the metric is changed to MeanSquaresImageToImageMetric.
Observed pretty much the same thing as (2) except the metric values are
different.

So what did I do wrong or where is the problem. Thanks very much in advance
for any help.

Ming
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20060106/94c81a78/attachment.html