The author provides an implementation of the paper by M. Chen that is intended to provide a faster and more accurate inverse displacement image filter.

The algorithm is faster and more accurate than the InverseDeformationFieldImageFilter.

More quantitative test cases and analysis are needed to verify that 1) the class is faster and 2) the class is more accurate.

Source code and input images are provided.  The images and deformation found in the article not included are distributed with ITK.

I compiled it with both ITK 3.20.0 and ITK4 git.  It compiled fine.  The test failed because of an incorrect filename in the CMakeLists.txt.  Only the most basic test code is included.  I also wrote code to run the warping and inverse warping on the input image to reproduce the output of the first example present in the paper.

The code was usable.

The code was easy to read.  There is a Win32 header included in the test case for unknown reasons.  Some unused code was still present.  In terms of ITK coding styles some changes could be made -- no tabs, bracket location, method definition style.

It was a couple of well chosen datasets.

Thanks for this nice contribution.

I have made some suggested changes that can be found here:

http://gitorious.org/itk-repos/itk-fixed-point-inverse-deformation/commits/thewtex

Matt McCormick (thewtex) (9):

      Initial import from IJ revision #1.

      Fix testing file names.

      Remove executable permissions.

      The test generates the images shown in the article.

      Replace indentation tabs with spaces.

      Add UseInputImageInformation method.

      rm unused TimeProbe header.

      rm unused WarpVectorImageFilter header.

      Make multi-threaded.