Due to several reasons, the spectral extraction from the extended sources is much more complicated than from point sources (where the local background subtraction is generally used).
The first reason is spatial vignetting which significantly reduces the background intensity for off-axis direction. The second reason is the different electronic noise in different parts of CCDs which may also taken into account. Due to these two effects, simple local background subtraction does not work.
Another reason is the non-homogeneous distribution of soft cosmic protons which can be interpreted as photons with lower energy (so they have different vignetting factors from those of instrumental background). In addition, this background component is extremely variable and thus its quiscent level should be estimated. The other reason is the spatial distribution of different instrumental lines, which originated presumably in places where corresponding element's concentration is higher [5]. In adition, these two reasons complicate the blank-sky background subtraction (because the blank-sky background contains a mixture of instrumental and soft proton background component, together with cosmic X-ray background component which depends on the sky position).
Finally, the other important variable background component is the solar wind charge exchange. Most important properties of different background components are briefly summarized in [1].
To handle with all this difficulties, the Extended Sources Analysis Software (ESAS) has been developed. The newest version of the software analyzes the data from all 3 CCD cameras for EPIC instrument on-board XMM-Newton. It also contains the specially developed software for spatial analysis. The ESAS manual is located at [2], see also [3, 4] for details. Now, ESAS is a part of SAS.
Among the negative sides of ESAS is its complexity. Like another complicated software, it should be thoroughly studied before its usage in a particular case. Nevertheless, it provides an important cross-check of self-developed methods and can be used without a hesitation to solution of problems similar to those specified by their authors, e.g. derivation of X-ray emission parameters from the sample of X-ray objects covering the large part of XMM-Newton Field-of-View (e.g. galaxy clusters).
It should be noted that for coded mask telescopes the spectral extraction from extended sources is even more complicated because it is not possible to determine the direction of upcoming particle. Several techniques can be used in this case, see e.g. [6, 7].