ISOCAM Interactive Analysis User's Manual Version 5.0 - ISO - ESA

280 CHAPTER 21. USING SLICE WITHIN CIA
The third one corresponds to the “inflat” method of CIA where you provide a flat-field to
SLICE. The only enhancement here is that you can either use a single flat-field image or provide
a flat-field cube with as many planes as the data cube. Assuming you have this flat in variable
my flat, to use this method you need to type:
CIA> red param=set red param(tdt=’65801627’)
CIA> act = set act(/make map)
CIA> slice pipe,flat in=my flat
For some reason, SLICE will erase your flat variable during its processing, so before starting
slice pipe, make a copy of your flat-field cube.
Now the three specifically SLICE methods are:
• Sliding Mean flat-field: where the flat-field is derived from the mean of the data cube
taken on a sliding window of readouts. This is satisfactory if your sources are not very
strong, but if this is not the case, or if you have large scale structures, you run the risk
of seeing some of these structures go into the flat-field, which is not advisable. You could
protect yourself from that using a large sliding window, typically around 10 time N exp but
this is not always possible (i.e. when you have a small number of raster positions).
• Perturbed Single Flat-Field: although this does not sound much better, this is actually
an improved flat-field method. Here, using a median filter on the individual raster pointings,
the object structures are removed before an “auto”-like flat is computed. Then, still
using a sliding window over the readouts, perturbations to this single flat-field are computed.
The interesting point here is that since the object structures have been removed,
the size of the window can be smaller than in the previous case.
• Variable flat with Sky Divided (or DivSky): this is the ideal method where an
estimation of the sky is divided out of the images before the flat-field is computed. This
estimate of the ideal sky is done by smoothing the current value of the map. Therefore
it is immediately clear than a first flat-field correction must have occurred through the
make map action 3 . This is also the reason why two successive make map actions using
DivSky with the same settings can produce different maps (although the differences are
small). This is however the method which can be affected by the long-term transient or
affect its determination. Therefore we recommend to use it after the long-term transient
has been removed, and in combination with another flat-field method to produce the first
version of map.
You will rapidly see that in general, as far as flat-fielding quality goes, the DivSky method
produces apparently the best results (compare, for instance, Fig. 21.2 with Fig. 21.3). However,
it is also the method whose interaction with the long-term transient correction is the most
complex, and which can lead to strong artifacts. Therefore, in this section we describe them
both, but in sec. 21.5.3 we will use the Perturbed Single Flat-Field method to determine
the long-term transient correction.
Also note that SLICE now gives you the choice of working either on the cube of raster
positions, or on the full readout cube. The latter option is obtained by adding /docube on the
slice pipe command line. Although this is in principle better, the differences at this stage are
3 try using it at the start of your SLICE session and you will see that SLICE uses the Sliding Mean
Flat-Field method instead!