At the moment the background is calculated by smearing the image with a function of the form
.
This function is related to the Mexican hat function through its Fourier Transform. It has the nice property that it weighs more the outside ring, so that possible sources do not give too much of a strong signal at the source position. For the parameter I choose to use a value twice the size of the wavelet scale under investigation (so for every scale a different background map is generated).
After one iteration the input for the background estimate is the original image minus the sources, detected at scales smaller or equal to the wavelet scale under investigation. In this way sources with a large extent are modelled in the background map. This prevents spurious detections of small sources within the envelope of extended sources.
XMM-Newton SOC/SSC -- 2014-11-04