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

21.6. BAD PIXELS, GHOSTS AND SOURCES 287
Figure 21.6: The results of the long-term transient correction and variable flat-field determination.
Variable flat-field was performed using the DivSky method, with parameter setup as
indicated in Table 21.4. LW3 is on the left, and LW2 on the right. Compare with Fig. 21.1 to
measure the improvement
21.6.1 Removing bad pixels
The method used to identify remaining bad pixels, while protecting the sources (a problem in
previous versions of SLICE is quite simple:
1. in the /docube option, the cube is flat-fielded and a local cube of raster positions is created.
2. a local copy of the sky map is created and is smoothed successively with a window of 5, 7
and size filter pixels.
3. This smoothed map is used to identify and remove ghosts in the flat-fielded datacube. If
the error map has not been provided (see below) then a mean error level is computed at
that stage.
4. Each plane of the cube is projected on the sky and, for the defined pixels, the difference
map is computed between that image and the smoothed map created above. We now have
a cube of residuals.
5. four sky maps are created summing the residuals that fall in the following intervals: above
a given positive threshold (bin 1), between 0 and this positive threshold (bin 2), between
0 and a given negative threshold (symmetrical from the positive one, bin 3), and below
this negative threshold (bin 4). The number of readouts that fall in these four bins is also
computed.
6. An empty sky map is created and for each sky pixels and: if more that two thirds of the
readouts fall in bins 3 and 4 (all readouts below the level in the smoothed map) the sky
value is taken from the readouts in bin 3; if more that two thirds of the readouts fall in
bins 1 and 2 (all readouts above the level in the smoothed map) the sky value is taken