Hi Andrea and Sander,
why do you subtract your background instead of dividing?
Because light pollution gradients are
additive effects: they accumulate during the whole exposure time. Vignetting, on the other hand, is a
multiplicative effect: it is a constant function that modifies the illumination distribution, independently on exposure time. For this reason, to correct for vignetting we multiply the image by the inverse of a special control frame that mimics the distribution of illumination on the focal plane.
Additive effects are in general impossible to predict, so the only way of removing them efficiently is to build and subtract a model of the illumination irregularities in the image. Fortunately, in deep-sky images we have a perfectly
constant reference: the sky background. Since the sky background is constant (read
flat), we can place samples on background areas of the image to build an accurate
background model, which reproduces the existing illumination irregularities. DBE and ABE are two sophisticated image processing tools to build background models in PixInsight.
Note that background modeling and correction of multiplicative or additive effects have nothing to do with linearity of the data.
doing DBE is not the same as flat fielding and IMO it does not replace it
Actually not the same, but only for practical/methodological reasons. In theory, a synthetic background model could be as good as a natural flat frame. In practice, however, a natural flat frame is always preferable for two main reasons:
- When building a synthetic model, we can't isolate additive from multiplicative effects when both are present, as usually happens.
- Making an accurate flat frame is in general very easy, while building an accurate synthetic model is a delicate task, and sometimes it's almost impossible. For example, when there are very few background areas, we can make just an approximation based on marginal data (which is much better than nothing, of course
).
