SMOOTH

        SMOOTH

    Smooths an input map  using  various   methods,   all   based   upon   a
    convolving  kernel  of  5  points large only.  This method is crude, but
    very fast,  and  supports  blanking  in  principle...   Data  cubes  are
    processed plane by plane.

      1.  BOX A simple 5 by 5 boxcar smoothing.  This is   a   very   strong
        smoothing...
      2.  GAUSS A gaussian smoothing.  The gaussian is sampled on a 5  by  5
        grid   so  it  must not be too broad.  3 pixels seem a maximum.  Use
        task  GAUSS_SMOOTH  if  you  want   to   smooth   with   a   broader
        gaussian,   but   beware   it   does   not  handle  properly blanked
        pixels.
      3.  HANNING Smoothing by a 5 by 5 pyramid, with weights 3 2 1.
      4.  USER User defined smoothing coefficients on  a  5   by   5   grid,
        assuming   biaxial   symmetry.    Hence   6   coefficients  only are
        required, S00 S10 S11 S20  S21  S22.   The  corresponding  smoothing
        kernel is :
                                S22 S21 S20 S21 S22
                                S21 S11 S10 S11 S21
                                S20 S10 S00 S10 S20
                                S21 S11 S10 S11 S21
                                S22 S21 S20 S21 S22
        Such a generalised  "smoothing"  enables  some   image   enhancement
        based   on   gradients   (e.g.    with   weights 4 -1 0 0 0 0).  The
        "smoothed" image is normalised by the sum of absolute values of  the
        coefficients.