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

21.5. A WORKED EXAMPLE 285
Figure 21.4: An example of artifacts obtained with a incorrect long-term transient determination:
the signal oscillates and the number of complete oscillations is roughly equal to half the number
of raster legs. The dashed line represent the fitted correction (see text for details). In fact, these
artifacts were generated while using the DivSky flat-field method in the long-term transient
determination for the LW3 image. The LW2 image shows similar problems.
the graph there are two curves, a continuous one and a dashed one. The continuous on is the
exact correction, as derived from the data. It will, by definition (see M. A. Miville-Deschênes’
paper), end with a zero value, i.e. no long-term transient correction for the last raster position.
The dashed one is a fitted correction. Most of the time, the fit and the exact solution apparently
disagree. This is because the so-called “fitted-correction” is not a fit to the exact correction,
but rather a correction derived assuming the long-term transient is the combination of two
exponentials whose parameters are fitted. Therefore, when these two corrections agree in shape,
you have a good sign that you have reached the “right” correction (an offset is not important as
it will be taken care of by your background subtraction). On the contrary, when they disagree,
it is a good indication that you should tune the parameters better.
Once you have reached a satisfactory determination, it can be subtracted from your data
using the commands suggested by SLICE. Note that in some very complex case, it may be
interesting to subtract the fitted correction first and then iterate flat-field determination and
long term transient correction with the exact determination for the remaining residuals.
If, for whatever reason, you have determined the long-term transient in any other way, and
stored it in a variable call my ltt, then the command to use to subtract it is (I admit it is a
bit complex and this is why the code itself prints it so that you can cut and paste it on the
command line):
CIA> c[*,*,im param.indfiltre] = la sub(c[*,*,im param.indfiltre], my ltt)
21.5.4 Second flat-field determination
In a number of cases, this will be the last one, producing your final image. In principle, at this
stage the main component of the long-term transient has been removed and it is safe to use the
DivSky method. However, remember that it is better to first reconstruct an estimate of the