[Insight-users] AffineTransform

[suresh]

Hi Luis,

Thanks a lot for help.It will be alot more helpful if you clarify 
a few things on registration with AffineTransform

1. What about the transform parametres. I observed(by calling 
transform->GetNumberOfParameters()) that there are 12 to be 
given.What are they?
2. Is this peice of code still valid?
   vnl_quaternion<double> 
quat(solution[0],solution[1],solution[2],solution[3]);
   vnl_matrix_fixed<double,3,3> vnlmat = quat.rotation_matrix();
3. I'm not going to swap the fixed and moving images(MRI, SPECT), 
but while Transformation i'll get the Inverse matrix and apply.Is 
that right?

This is modified code. for registration.

 	   Matrix* mat = NULL;
 	   double result[16];

 	   typedef BufferToImageConversion<unsigned char> 
ConverterType;
 	   typedef ConverterType::ImageType UCharImage;

 	   typedef itk::AffineTransform<double> TransformType;
 	   typedef itk::RegularStepGradientDescentBaseOptimizer 
OptimizerType;
 	   typedef itk::LinearInterpolateImageFunction <UCharImage, 
double> InterpolatorType;
 	   typedef itk::ImageRegistrationMethod <UCharImage , UCharImage 
 > RegistrationType1;

 	   typedef itk::MutualInformationImageToImageMetric<UCharImage , 
UCharImage> MetricType;
 	   ProgressUpdate(50, "Performing MutualInformation 
Registration", MRIVolumeName);
 	   MetricType::Pointer         metric        = 
MetricType::New();
 	   TransformType::Pointer      transform     = 
TransformType::New();
 	   OptimizerType::Pointer      optimizer     = 
OptimizerType::New();
 	   InterpolatorType::Pointer   interpolator  = 
InterpolatorType::New();
 	   RegistrationType1::Pointer  registration  = 
RegistrationType1::New();

 	   ConverterType converter;
 	   UCharImage::Pointer fixedImage = converter.GetImage(MRI);
 	   UCharImage::Pointer movingImage = 
converter.GetImage(SPECT);

 	   RegistrationType1::ParametersType 
guess(transform->GetNumberOfParameters() );

 	   CString str;
 	   str.Format(" Count %d", 
transform->GetNumberOfParameters());
 	   AfxMessageBox(str);

 	   guess[0] = 0.0;	guess[1] = 0.0;	 guess[2] = 0.0; guess[3] = 
1.0;
 	   guess[4] = 0.0; guess[5] = 0.0; guess[6] = 0.0; guess[7] = 
0.0;
 	   guess[8] = 0.0; guess[9] = 40.0;   guess[10] = 20.0; 
guess[11] = 0.0;
 	   try{
 		   registration->SetInitialTransformParameters (guess);
 		   metric->SetMovingImageStandardDeviation( 2 );
 		   metric->SetFixedImageStandardDeviation( 2 );
 		   metric->SetNumberOfSpatialSamples( 1000 );
 		   typedef OptimizerType::ScalesType ScaleType;
 		   ScaleType scales(transform->GetNumberOfParameters());
 		   scales.Fill( 1.0 );
 		   for( unsigned j = 4; j < 12; j++ )
 		   {
 			   scales[j] = 1.0 / vnl_math_sqr(300.0);
 		   }
 		   //Scale parameters
 		   optimizer->SetNumberOfIterations( 7 );

 		   //set registration parameters
 		   registration->SetMetric(metric);
 		   registration->SetOptimizer(optimizer);
 		   registration->SetTransform(transform);

 		   registration->SetInterpolator(interpolator);

 		   registration->SetFixedImage(movingImage);
 		   registration->SetMovingImage(fixedImage);
 		   registration->SetFixedImageRegion( 
fixedImage->GetBufferedRegion() );
 		   // Setup the optimizer
 		   optimizer->SetScales(scales);
 		registration->DebugOn();
 		   registration->StartRegistration();

 		   ProgressUpdate(60, "Performing Normalized Registration15", 
MRIVolumeName);
 		   // Get the results
 		   itk::Array<double>  *arr=new itk::Array<double>;
 		   RegistrationType1::ParametersType solution 
=registration->GetLastTransformParameters();

 		   vnl_quaternion<double> 
quat(solution[0],solution[1],solution[2],solution[3]);

 		   vnl_matrix_fixed<double,3,3> vnlmat = 
quat.rotation_matrix();

 		   result[0] = vnlmat[0][0]; result[1] = vnlmat[0][1];
 		   result[2] = vnlmat[0][2]; result[3] = solution[4];
 		   result[4] = vnlmat[1][0]; result[5] = vnlmat[1][1];
 		   result[6] = vnlmat[1][2]; result[7] = solution[5];
 		   result[8] = vnlmat[2][0]; result[9] = vnlmat[2][1];
 		   result[10] = vnlmat[2][2]; result[11] = solution[6];
 		   result[12] = 0;	result[13] = 0;	result[14] = 0; result[15] = 
1;

With this code i could not go beyond the StartRegistration call? 
Please help  in fixing this..

And in Resample code i'm using the Inverse of Transform.
 	TransformType::Pointer  inverseTransform = 
transform->Inverse();
 	resampleFilter ->SetTransform(inverseTransform);

Thanks
suresh