User:Ramirez: Difference between revisions

From KitwarePublic
Jump to navigationJump to search
No edit summary
mNo edit summary
Line 49: Line 49:
In particular, look at lectures 8 and 9.
In particular, look at lectures 8 and 9.


[...] It is not productive to fight with the registration + optimization
[...] It is not productive to fight with the registration   optimization
until you find a way of generating relatively smooth Metric plots.
until you find a way of generating relatively smooth Metric plots.
Note that this is just an exercise on learning how to tune the
Note that this is just an exercise on learning how to tune the
Line 96: Line 96:
     movingReader->Update();
     movingReader->Update();
     }
     }
   catch( itk::ExceptionObject & excep )
   catch( itk::ExceptionObject
    {
    std::cerr << "Exception catched !" << std::endl;
    std::cerr << excep << std::endl;
    }
 
 
  typedef itk::MattesMutualInformationImageToImageMetric< ImageType, ImageType >  MetricType;
 
  MetricType::Pointer metric = MetricType::New();
 
 
 
  typedef itk::TranslationTransform< double, Dimension >  TransformType;
 
  TransformType::Pointer transform = TransformType::New();
 
 
 
  typedef itk::LinearInterpolateImageFunction<
                                ImageType, double >  InterpolatorType;
 
  InterpolatorType::Pointer interpolator = InterpolatorType::New();
 
 
  metric->SetInterpolator( interpolator );
  metric->SetTransform( transform );
 
  metric->SetNumberOfHistogramBins( 20 );
  metric->SetNumberOfSpatialSamples( 10000 );
 
  transform->SetIdentity();
 
  ImageType::ConstPointer fixedImage  = fixedReader->GetOutput();
  ImageType::ConstPointer movingImage = movingReader->GetOutput();
 
  metric->SetFixedImage(  fixedImage  );
  metric->SetMovingImage( movingImage );
 
  metric->SetFixedImageRegion(  fixedImage->GetBufferedRegion()  );
 
  try
    {
    metric->Initialize();
    }
  catch( itk::ExceptionObject & excep )
    {
    std::cerr << "Exception catched !" << std::endl;
    std::cerr << excep << std::endl;
    return -1;
    }
 
 
  MetricType::TransformParametersType displacement( Dimension );
 
  int rangex = 50;
  int rangey = 50;
 
  for( int dx = -rangex; dx <= rangex; dx++ )
    {
    for( int dy = -rangey; dy <= rangey; dy++ )
      {
      displacement[0] = dx;
      displacement[1] = dy;
      const double value = metric->GetValue( displacement );
 
      std::cout << dx << "  "  << dy << "  " << value << std::endl;
      }
    }
 
  std::cout << std::endl;
 
 
  return 0;
}</pre>
 
 
http://public.kitware.com/pipermail/insight-users/2004-April/008043.html
 
 
 
 
<b><u>Optimizer->SetScales()</u></b>
 
 
The rule of thumb is to figure out how much each one of those
parameters will change for your registration, and then rescale
that range to [-1:1].
 
 
 
<u>In the case that you know the anticipated range of translations and rotations,</u>
 
 
"[...]if you are doing 2D rigid you will have a 2D transform with
three parameters:
 
      Tx  translation in millimeters along X
      Ty  translation in millimeters along Y
      R  rotation in radians
 
and you anticipate that your images need a correction of the
order of 10 to 50 millimeters in translation and 0.01 to 0.1
radians in rotation, then you should put scales:
 
      scale[0] = 1/50;    scale for Tx
      scale[1] = 1/50;    scale for Ty
      scale[2] = 1/0.1;    scale for Rotation
 
Of course, those will be just "good values to start with".
You will still need to refine them according to the behavior
of the optimizer."
 
http://public.kitware.com/pipermail/insight-users/2005-April/012896.html
 
 
 
<u>In the case that you do not know the anticipated range of translations and rotations,</u>
 
 
"[...]the recommendation for the scaling of translation parameters versus
rotation parameter is to use a factor proportional to the diagonal
length of the image.
 
For your case the, you have 100 pixels with 1 mm / pixel, therefore the
physical extent of your image is
 
        100mm  X  100mm  X 100mm
 
The diagonal the image bounding box is
 
          sqrt(3) * 100 mm
 
which is about
 
              173.2
 
and extra factor of 10X is usually useful, so you should probably try a
factor of
 
    1.0 / ( 10 x 173.2 )  =  1.0 / 1732.0
 
You could use this same factor for the three components
of the translation or you could estimate independent
factor for each component in the way it is done in the
VolView plugin.
 
Note that this factors are not expected to be computed precisely. Their
purpose is simply to bring the rotational and translational parameters
to a similar numerical scale.
 
By default, they are quite disproportionate since rotation
are in radians, therefore in a range about -1:1, while translations are
in millimeters, and for an image of 100mm you probably can expect
translations as large as 50mm."
 
http://public.kitware.com/pipermail/insight-users/2004-July/009558.html
 
 
 
In short,
 
 
"[...]for an 3D AffineTransform, you get 12 parameters: the
first 9 are the coefficients of the matrix (representing
rotation, scale and shearing) the last 3 are the components
of a translation vector.  You want then to provide an
array of 12 values with the first 9 being =1.0 and the last
three being on the range of 1.0 / the image size (in millimeters)."
 
http://public.kitware.com/pipermail/insight-users/2002-October/001400.html
 
 
 
<b><u>itk::RegularStepGradientDescentOptimizer->SetMaximumStepLength()</u></b>
 
<b><u>itk::RegularStepGradientDescentOptimizer->SetMinimumStepLength()</u></b>
 
 
"[...]There is no magic recipe for selecting one. You probably
want to start experimenting with a small value (e.g. 0.01)
and plot the metric evaluations during the registration
process.  If you observe that the metric values are fairly
monotonic, that means that you can safely increment the
step length. Such an increment has the advantage of reducing
the time required to reach an extrema of the cost function
(the image metric in this case).  You could restart the
registration with larger values of the step length, as long
as you don't observe a noisy and/or erratic behavior on the
Metric values.
 
Step length issues are discussed in the course material
from the "Image Registration Techniques" course at RPI.
 
  http://www.cs.rpi.edu/courses/spring04/imagereg/
 
for example in lecture 9:
 
  http://www.cs.rpi.edu/courses/spring04/imagereg/lecture09.ppt"
 
 
http://public.kitware.com/pipermail/insight-users/2004-July/009558.html
 
 
 
 
== Use Cases ==
 
 
=== CT-MRI Brain Registration ===
 
 
 
=== PET-CT Registration ===
 
 
 
{{ITK/Template/Footer}}

Revision as of 00:30, 18 April 2007

Image Registration Components

Image Similarity Metrics


Transforms

Optimizers

Interpolators

Tuning Parameters

Optimizers

"[...] In order to get some insight on how to tune the parameters for your optimization, probably the best thing to do is to study the characteristics of the Metric when computed over your images. Please find attached a small program that allows you to explore the values of the metric using a translation transform. You want to plot these values in order to get a feeling on the level of noise of the Metric, the presence of local minima, and the overall reproducibility of the Metric function itself. (you will have to convert the dimension to 3D, since the example was configured for 2D)

Please run this program first using the *same* input image as fixed and moving images, and let the translation be evaluated for a range that is close to 1/2 of the image extent (physical extent in millimeters).

Plot the values using your favorite plotting program. (anything from GNUPlot to Excel).

For examples on how these plot will look like, please take a look at the course in Image Registration: 

    http://www.cs.rpi.edu/courses/spring04/imagereg/

In particular, look at lectures 8 and 9.

[...] It is not productive to fight with the registration optimization until you find a way of generating relatively smooth Metric plots. Note that this is just an exercise on learning how to tune the parameters. Once you figure out the parameters, you will not need to plot the Metric anymore.

Make sure that origin and spacing are correctly set in your images before you start computing all these metric plots." 


#include "itkImage.h"
#include "itkImageFileReader.h"
#include "itkImageFileWriter.h"
#include "itkMattesMutualInformationImageToImageMetric.h"
#include "itkTranslationTransform.h"
#include "itkLinearInterpolateImageFunction.h"


int main( int argc, char * argv[] )
{
  if( argc < 3 )
    {
    std::cerr << "Usage: " << std::endl;
    std::cerr << argv[0] << "  fixedImage  movingImage" << std::endl;
    return 1;
    }

  const     unsigned int   Dimension = 2;
  typedef   unsigned char  PixelType;

  typedef itk::Image< PixelType, Dimension >   ImageType;
  typedef itk::Image< PixelType, Dimension >   ImageType;


  typedef itk::ImageFileReader< ImageType >  ReaderType;

  ReaderType::Pointer fixedReader  = ReaderType::New();
  ReaderType::Pointer movingReader = ReaderType::New();

  fixedReader->SetFileName(  argv[ 1 ] );
  movingReader->SetFileName( argv[ 2 ] );

  try 
    {
    fixedReader->Update();
    movingReader->Update();
    }
  catch( itk::ExceptionObject
Retrieved from "https://public.kitware.com/Wiki/index.php?title=User:Ramirez&oldid=8799"