New in Release 1.11.0
---------------------
* Make nu_estimate_np_and_em.in invariant to voxel steps (via
  spline_smooth in EBTKS - must use EBTKS 1.6.2 or above)
* Free memory after usage
* Remove downsampling in evaluation of field (computers are fast
  now, so be more accurate by sampling at full resolution)
* Correct the field in the region outside the mask to make
  sure the mask varies smoothly away from the brain (important
  with the previous changes)
* New program for the far-field correction

New in Release 1.10.1
---------------------
* Simply repackage the tar ball with good m4 files for minc2
* Changed xPERLx to get good perl exec

New in Release 1.10
-------------------
* Bug fixed in nu_estimate_np_and_em.in that occasionally caused mincmath
   to be called with the same filename for both an input and an output.
   Among other things, this sometimes caused test #14 to fail.
* Minor change to extracttag.c to avoid problems on 64-bit platforms.
* Added checks for finite/isfinite for Solaris

New in Release 1.09
-------------------
* Build fixes.  No user visible changes.

New in Release 1.08
-------------------
* Fixed make_template script to avoid crash commonly seen with files
   containing lots of DICOM data in their headers.
* Build of PERL scripts fails when using IRIX make.  The result is a set of
   scripts of zero lengths.  The workaround is to build with GNU make instead
   of the default IRIX make.

New in Release 1.07
-------------------
* Build fixes.  No user-visible changes.

New in Release 1.06
-------------------
* Improvements to build procedure: configure detects FORTRAN LAPACK or CLAPACK.

New in Release 1.05
-------------------
* The functionality of release 1.05 is unchanged from that of 1.04.  
* A number of minor changes have been made to the code, Makefiles, and
   configuration script to allow it to build with gcc 2.95.  A number
   of additional changes have been made which (I hope) will allow it
   to compile on Solaris machines without complaint.

New in Release 1.04
-------------------
* The most important change with 1.04 is that the code can now be compiled
   with the new egcs version (1.0 or later) of g++.  This means the current
   distributions of Linux, most of which come with egcs, should now be able to
   compile N3.
* Release 1.04 fixes a number of bugs from release 1.02.  Among these,
   a change to the way the .imp correction field is produced at the final
   stage means that the new version produces slightly different results than
   previous versions (typically by ~ 0.3%).  Also, MNI_AUTOREG is no longer
   needed to install N3.

New in Release 1.03
-------------------
* A new verison of the test data is needed for the quantitative tests.
  > fixed 1.04
* float volumes created by evaluate_field may suffer round off problems
  > fixed in 1.04
* mincblur is required but not really necessary most of the time
  > fixed in 1.04

New in Release 1.02
-------------------
* volume_hist crashes if image min and max are same
   > fixed in 1.02c
* nu_correct fails rather mysteriously if the mask volume is empty
   > fixed in 1.04
* The code for bining in volume_hist may not handle the last bin correctly. 
* The imp generation stage of nu_estimate_np_and_em may fail due to under
   sampling of smooth volume.
   > fixed in 1.03.   Note: this fix changes the results of the algorithm
   slightly ( less than 0.3% for test suite ).
* The intensity scaling of the output volume can be quite different than the
   input.
   > While this isn't a bug, an option has been added to nu_estimate_np_and_em to
   correct for this in version 1.03.
   > This option has been added to the nu_estimate/nu_correct scripts as of 1.04

New in Release 1.1
------------------
* Path.c doesn't work correctly with the bash shell since it doesn't
   have a PWD variable.
   > fixed in release 1.04
* On SGIs the compiler should be CC not DCC to take advantage of the
   new formats.

New in Release 1.0
------------------
* The make_template command assumes the MINC header will be less than 16k.
   This may cause evaluate_field to crash if template header has been
   truncated.  
   > changed to 32k in release 1.02
* mv command is not registered in nu_evaluate script.
   > fixed in release 1.02
* Software doesn't compile on g++ versions that don't support -O3.
* spline_smooth doesn't check that output file is writable until after
   computing field.
* evaluate_field may compute a different number of basis functions than 
   spline_smooth did as a result of round off errors.
