Revision 1
__________

Platforms
---------

- Tested under Windows XP MSVC7.1 
- Compiles also in Linux gcc 4.1.1

class GaussianDerivativeOperator
--------------------------------

- Scale-space normalization and use of image spacing must be checked again.

class DiscreteGaussianDerivativeImageFunction
---------------------------------------------

- The final N-D kernel size in each dimension (calculated by convolution) is set to twice the original size for that dimension. 
  Maybe this size could be reduced without losing precision.
- There are too many RecomputeGaussianKernel() calls. Maybe it would be better to set a modification flag that check if the
  parameters have been modified. This flag would be check in the first call to EvaluateAtIndex(), that would call
  RecomputeGaussianKernel().
  
class DiscreteHessianGaussianImageFunction
------------------------------------------

- Computation of derivatives could be further speed up as it is a bit slow now

Batch files
-----------

The *.bat files contain the command line paramaters used in the paper examples


Revision 2
__________

- Fixed some errors with std::stringstream and itk::ImageFileReader

Revision 3
__________

- Changed wrong comment for member m_OperatorArray in in itkGaussianDerivativeImageFunction.h .
- Added option to choose nearest neighbour or linear interpolation through member m_Interpolation mode in 
  itk::DiscreteGaussianDerivativeImageFunction and itk::DiscreteHessianGaussianImageFunction.
- Fixed wrong template macros in itkDiscreteGaussianDerivativeImageFunction.h and 
  itkDiscreteHessianGaussianImageFunction.h.
- Added new class itkDiscreteGradientMagnitudeGaussianImageFunction.h
- Added typedefs for PointType and ContinuousIndexType in image functions.
- Added new examples InputImageDerivatives and InputImageDerivatives3D to calculate arbitrary derivatives of an input image.
- Fixed error in GaussianImageDerivatives3D.cxx the order set for the z direction operator was wrong.
- Added Tests using GaussianImageDerivatives.
- Fixed bug in itk::DiscreteHessianGaussianImageFunction. Order of derivatives in the Hessian matrix was incorrect.