A robust algorithm for sky background computation in CCD images

In this paper we present a non-interactive algorithm to estimate a representative value for the sky background on CCD images. The method we have devised uses the mode as a robust estimator of the background brightness in sub-windows distributed across the input frame. The presence of contaminating objects is detected through the study of the local intensity distribution function and the perturbed areas are rejected using a statistical criterion which was derived from numerical simulations. The technique has been extensively tested on a large amount of images and it is suitable for fully automatic processing of large data volumes. The implementation we discuss here has been optimized for the ESO-FORS1 instrument, but it can be easily generalized to all CCD imagers with a sufficiently large field of view. The algorithm has been successfully used for the UBVRI ESO-Paranal night sky brightness survey (Patat 2003).