can't do DBE on this image

[pfile]

i don't know what's special about this particular image, but it is the first time i've captured data thru an OIII filter on my DSLR.

the image is the green channel only from the result of the integration of 3 (poorly guided) lights. the lights are only partially calibrated - i have not had time to make flats thru the OIII filter yet. they were calibrated with a master bias and master dark frame, with dark calibration and optimization turned on.

no matter what i set the tolerance to (all the way up to 10) or the shadow relaxation, i can not get even one good sample on this image. no idea what's going wrong but it would be great if someone could take a look at it and let me know what i'm doing wrong, or if in fact this image has exposed some weirdness in DBE. i was able to use ABE to sort of get rid of the vignetting and gradients but the background model was not very accurate.

thanks.

http://dl.dropbox.com/u/8884840/integration1_G.fit.zip

edit: i just discovered that if i crop off the black areas then i can automatically or manually create working samples with "sane" tolerances. is this expected behavior? also if i save the process icon for the successful DBE, undo the crop on the original image, and then reapply the DBE icon, all sample points that do not fall along the black area are red.

[Juan Conejero]

Hi Rob,

The image has very dark --but not black-- borders that are tricking DBE. Apply the following PixelMath expression:

iif( $T < 1e-2, med( $T ), $T )

with Rescale result disabled. This will replace those borders with the median of the image. Now DBE works without problems if you raise its tolerance parameter to 1.5 or so.

DBE's background modelling routines are robust but these borders have put it to its knees... I'll try to improve it.

[pfile]

okay, thanks. it's pretty common to have these kinds of borders when you point your mount by hand night after night...

[Nocturnal]

Well yeah but it's also common to first crop your image before processing it All kinds of processes are pretty meaningless if there's a border or region with very different background.

What I do:

- load image
- STF - A
- crop
- STF - A
- proceed

Quite a while back I actually asked Luc to add a new 'intersection' mode to DeepSkyStacker that automatically crops the image to the region where all subs contributed. It leaves a single pixel border that I crop in PI before proceeding.

[pfile]

in this case it's narrowband data, and i wanted to do DBE separately on each channel first. if i want to crop the image, i'd have to do an RGB merge, crop, then extract the channels again and do the DBE, and then re-merge into RGB. window clutter city. it seemed easier to do the DBE first and then the other stuff next.

ordinarily i'd be starting with RGB since i'm using a DSLR, but this is a new situation for me.

the intersection mode wouldnt really help here since i'm working with two separate stacks that are of course registered to the same light.