Here's what I've done to these images:

  o Flipped axes and swapped bytes to create the FITS data segments

  o Converted all image values less than 1.e-06 to zero.
      The MISSING_CONSTANT in the labels never occured in any of the images,
      but in checking the values in the navigation images, there was a jump
      discontinuity in absolute values from the black areas to the image
      areas, which all had magnitude of 1.e-03 or greater.  I'm guessing the
      MISSING_CONSTANT, which equates to a very large (e+36 or so) number,
      was a null value inserted prior to processing which was then processed,
      resulting in the varying small values in the black areas of the 
      navigation images. 

      Spot-checking the calibration images did not turn up any values with 
      magnitude smaller than about 1.e-03.

  o Collected values from the existing IMG files for the FITS header segments.
    Keywords added to FITS files:

      MISSION   set to 'STARDUST'
      INSTRUM   set to 'NAVCAM'
      DATE-OBS  set to START_TIME
      TARGET    set to TARGET_NAME
      EXPOSURE  set to EXPOSURE_DURATION
      RA        set to RIGHT_ASCENSION
      DEC       set to DECLINATION
      WAVELEN   set to 698.8

  o Converted attached labels to detached labels.

  o Omitted keywords from PDS labels for various reasons:

      MAXIMUM    omitted because the value is not trusted and annoying to check
      MINIMUM    omitted because the value is not trusted and annoying to check
      CHECKSUM   omitted because it is no longer valid
      MEAN       omitted because the value is not trusted and annoying to check
      STANDARD_DEVIATION omitted because the value is not trusted and annoying to check

  o Added UNIT and AXIS_ORDER_TYPE to PDS label.

  o Changed MISSING_CONSTANT to zero in PDS labels.

  o Replaced single quotes around string values in PDS label with double quotes.

  o Added HEADER object to PDS label.