DOC: Improve/correct TransformGeometryFilter documentation.

Modules/Core/Transform/include/itkTransformGeometryImageFilter.h

@@ -26,28 +26,29 @@
 namespace itk
 {
 /** \class TransformGeometryImageFilter
- * \brief Pseudo-Resample an image by modifying the geometric meta-data and not the pixels.
- *
- * The TransformGeometryImageFilter will generate a physical memory-modified version of
- * the input image if the input transform is not identity. Its neglectful use
- * can be a source of problems: e.g. it can exhaust the memory if the image is
- * very large, and it WILL reduce the image quality when there are lots of
- * transforms to be superimposed for the input image. Often times, we are not
- * interested in the intermediate transformed images.
- *
- * If all the transforms are rigid, there is a far superior way to achieve a similar result.
- * Updating image metadata in-place removes the accumulated resampling errors
- * as well as eliminating the expense of accessing the physical memory of the image.
- * We need to compose all the transforms beforehand to make use of this filter.
- *
- * \param \c RigidTransform Currently must be a VersorRigid3DTransform.
+ * \brief Modify an image's geometric meta-data, changing its "physical" extent.
+ *
+ * The TransformGeometryImageFilter "physically" changes the image in
+ * space using the given transformation. The
+ * specific transformation type can be any type derived from the
+ * MatrixOffsetTransformBase and the TranslationTransform.
+ * The modification of the geometric meta-data is an alternative to
+ * resampling the moving image onto the fixed image grid, after registration. The advantages of using
+ * this approach over resampling are two-fold, it does not introduce artifacts
+ * into the result because the original intensity information is not
+ * modified, and it is computationally more efficient.
+ *
+ * When the filter is used with a rigid or translation transformation
+ * the resulting image can be saved in any desired format. When the
+ * filter is used with an affine transformation the resulting image
+ * should be saved in a format that supports a non ortho-normal
+ * direction cosine matrix (e.g. nrrd).
+ *
+ * \param \c Transform Any transform derived from MatrixOffsetTransformBase or TranslationTransform.
  * \param \c InputImage The image to be duplicated and modified to incorporate the
- * rigid transform.
+ * transform.
  * \return An image with the same voxel values as the input, but with different
- * physical space representation affected by the rigid transform.
- *
- * The purpose of this class is to generate the new origin and direction
- * that will remove the need for using the transform.
+ * physical space representation affected by the transform.
  *
  * Let us call the transform operation from the fixed image to moving image <tt>TfmF2M</tt>.
  * Given a set of points from the fixed image in physical space (i.e. <tt>physicalFixedImagePoints</tt>),