Perhaps there is some logic to the contrary.... but I would recommend doing the background neutralization and color calibration before DBE.
I am curious, why do you think that it is better to do BN and CC before DBE? In my experience color calibration works better after removing the gradients of the image.
Andres,
BN and CC attempt to work from a reference of some sort (as does PCC using a star catalog). The idea is to take an image and remove offsets due to the differing sky brightness in each channel and then applying coefficients to each channel (globally, across the entire image, every pixel). So for BN you choose a bit of sky (a reference) and apply non-local (global). Same is true for CC where the coefficients are computed. If, you select a white reference such as a galaxy- this determines your coefficients for the white balance ratios. Using bright sources (structure method) is a different method that also will find the ratios.
DBE models changes to the background (and possibly more, depending on your tolerances) based on many local samples. This model could influence the above. Lets say for a given preview location for BN- I can certainly get multiple results at this location if I do DBE first- it depends on what I did in DBE that will determine my measurement for BN. There are no constraints on the DBE model.
Now you might argue that if you do CC before DBE and you choose that little spiral galaxy in your picture- the gradient will impact the result. This is very true. And, as I am guessing, it might lead you to say "Ah! I need to do DBE first." My response is to simply choose a method that makes the gradients small compared to the measure you are making. Structure detection is a good choice as you would expect significant structures are brighter than gradients. Let me be clear- if the statement that gradients are not much smaller than significant structures- then I fear this is an acquisition issue (glare, bad flats...etc etc). PCC is exactly the answer by analyzing stars which not be influenced in any significant way by gradients (given the assumption above).
So, BN and CC first (as well as PCC) because the reference if valid applies globally to the image- it is a constraint. DBE is unconstrained based on the great number of free parameters (choice of samples, tolerance..etc)- so it should be applied afterwards. I basically see BN+CC as a linear operation- whereas DBE is not.
-adam
