Next: Input Output data Up: ASMOOTH Prev: ASMOOTH
Algorithm
The ASMOOTH algorithm differs from that used in most adaptive smoothing
programs, which typically smooth with a running box whose size is
adjusted from point to point to enclose a constant number of counts.
ASMOOTH partitions the data range ("intensity" in the most common case)
into user specified slices. Within each slice a simple smooth is
performed, but the width of the smoothing function is chosen in each
slice to achieve the target s/n value, on average, within that slice.
i.e. (mean data value)/sqrt(smoothed mean variance)=target s/n.
Hence higher intensity slices will be smoothed progressively less.
The program informs the user of the width (or FWHM in the Gaussian
case) of the mask being used at each data level.
This approach typically produces a more visually attractive result
than conventional adaptive smoothing, since (i) the filter does not
need to be a block filter, (ii) the data slicing approach reduces
the leakage from very bright point sources, especially if log spacing
is chosen for the data slices, and (iii) the treatment of negative
counts amounts to a "total smooth". The disadvantage is that since the
processing is genuinely adaptive to the data, it is not possible to
propagate errors, or to repeat exactly the same adaptive smooth on different
datasets.
Next: Input Output data Up: ASMOOTH Prev: ASMOOTH