$\star$ The chart of satellite

To identify the satellites, the fact is used that they touch a great number of contiguous pixels and that their elongation is significant (in general, a satellite with a width of ten pixels and a length which can reach a length of CCD).

One thus seeks the pixels whose flux is with the top of a certain threshold (typically $2\sigma_{ciel}$), then one looks in an iterative way if the 4 close pixels are above the threshold. Gradually, one builds all the `` islands '' with the top of the `` sea '' of noise.

This procedure thus will select all the wide objects of the image and in particular stars and the galaxies.

A series of cut is applied to select only the objects having a significant elongation (typically a report/ratio large axis on small axis higher than 2), a large relatively long axis (of the order of the tenth of the diagonal of a CCD). Moreover, one asks so that the number of touched pixels be sufficiently large (of the order of the number of pixels over the width of the CCD). Lastly, to avoid introducing the dead pixels or the pixels of the saturated objects, one provides to the procedure the charts dead saturation and pixels which are not considered at the time of the procedure.