One necessary step of the data processing is the subtraction of the background. One way to do it is to create an average or a median image of the whole file series and then subtract each single image by this average/median. This method has the advantage that you not only get rid of the background noise but also all those reflections which are visible homogeneously in every image, such as pressure media, artefacts on the detector or some diamond spots.

Creating an average image can be done with timelessMeanFileSeries which is part of the TIMEleSS tools. Creating the median can be done with median.py which comes with Fabian.

The median image is less affected by outliers than the average. Therefore, if you have strong diamond peaks in some of the images, you will probably still see them also on the average image while there are none in the median one. Because of that, for subtracting the background the median image should be preferred. However, for other purposes you might still need the average image.

There are several options to subtract the background. It depends on your data and your goal which way you use. But whichever way you choose, only subtract the background once!