2 Sources detection

Object detection was performed across the area observed in optical bands (WFPC2 field) using the software for automatic sources detection SExtractor (Bertin & Arnouts 1996). The software firsts smoothes the images and then applies a threshold to identify peaks. As a smoothing function we have chosen a Gaussian function having FWHM equal to the one measured on the images (i.e. 0.16 arcsec) and a detection threshold of 1 sigma per pixel, with a minimum detection area equal to the seeing disk, have been adopted to pick up objects. The software allows to examine every identified group of contiguous pixels to deblend overlapping sources: two sources belonging to the same group are considered different objects if they differ of 5 magnitudes or less (deblend-mincont = 0.01, deblend-ntresh = 32).

This choice of low thresholds aimed at obtaining a raw catalogue not biased against very faint or small source, in fact the first sample consisted of 6093, 4747, 9850, 5229 detections in the F300W, F450W, F606W and F814W band respectively. From this sample we then extracted a catalogue optimizing completeness while minimizing the number of spurious detections by selection criteria based on the signal-to-noise ratio of sources and simulations (as explained in Sect. 4).