1 MAKE_CUBE
        MAKE_CUBE

    This is an image construction task  which   is   able   to   produce   a
    filled image from one containing many blanked pixels.  The reconstructed
    filled image is not constrained to fit exactly the observed data points.
    On  the  opposite, the construction is made by the analogy to a flexible
    plate attached to fixed points by springs:  the plate is  the  analogous
    of  the  surface  represented by the image, and the fixed points are the
    analogous of the observed data points.  By adjusting the parameter P
                          P = (plate stiffness) / (springs stiffness)
    it is possible to control the fidelity to  the  original  data  and  the
    amount of smoothing involved in the image reconstruction.  Low values of
    P mean  high  fidelity  to  observed  data,  and  negligible  amount  of
    smoothing.

    The original grid is first expanded  by   a   factor   EXPANSION$,   new
    pixels  being  attributed  the  blanking  value.  Then, the minimization
    proceeds iteratively to adjust the final  image,  until  convergence  is
    reached.   Initially  blanked  pixels  are  ignored  in  the convergence
    criterium.

    The algorithm works on cubes, processing each plane  independently.   It
    can  be  used as an alternative to FILL_CUBE for oversampled images, but
    works also for undersampled data.
2 P$
        TASK\REAL "Smoothing parameter" P$

    This is the value of the  smoothing  parameters.    It   should   remain
    between  0.001  and  10.  The higher values give stronger smoothing, and
    should be used for noisy data.
2 NITER$
        TASK\INTEGER "Number of smoothing iterations" NITER$

    This is the number of  smoothing   iterations.    Ten   iterations   are
    usually  enough  to  reach  convergence,  unless the expansion factor is
    large.
2 EXPANSION$
        TASK\INTEGER "Expansion factor" EXPANSION$

    This is the expansion factor for the output cube, i.e.  the   number  of
    output   pixels   for   each  input pixel along each direction.  Minimum
    value is 2, maximum is 10.  Note that the computing time is proportional
    to the number of output pixels, thus to EXPANSION$**2.
2 IN$
        TASK\FILE "Input file name" IN$

    This is the name of the input file.
2 OUT$
        TASK\FILE "Output file name" OUT$

    This is the name of the output smoothed image.
2 GUESS$
        TASK\REAL "Initial value for blanked pixels" GUESS$

    This is an initial guess  for   the   original   value   attributed   to
    blanked  pixels.   Try to find something reasonable, usually the average
    of valid pixels.  Convergence can be  delayed  if  you  choose  a  crazy
    value,  since  the minimization tends to produce a smooth image, but the
    result doesnot depend on GUESS$.
1 END