DynamicBackgroundExtraction Execution Cycles

[smccully]

DBE is very useful tool for flattening an image, and removing gradients. Its hardly ever a single execution. In severe cases its very often that running DBE > 5 - 10 times on a single image. Even more so its also can be very much a trial and error process, in trying to create a model that removes the graidents and preseves details.

So after running DBE several times, and then undoing those changes to revert the image back to its initial state it seems useful that you could execute the DBE Model N times, it may even make sense to execute the model N times where N produces some diminshed returns. This would be where continuing to execute DBE no longer produces a meaningful change to the image.

To try and summarise,

1. Include a cycle to a DBE model, where DBE will execute the model N times and/or allow a model to be executed until any continued executions no longer produce any meaningful change.
2. This allows for a much easier and even more automated workflow
3. Allows an easier iteration of the DBE model, which allows you to undo one change versus multiple changes and cycling through swap files. Which when making very large mosaics can be difficult and computational intense process.

 

[smccully]

I think the same could also apply to AutomaticBackgroundExtractor as well.

 

[pfile]

10 times! do you mean for testing the control points or actually applying DBE to the image 10 times to remove the gradient?

i don't think i've ever had to do more than 2 or 3 times (and 3 would be a real outlier) even for my worst gradients.

 

[dld]

10 times sounds way too much. I use wide-angle lenses with lots of vignetting and even for sessions where the lens sweeps a large percentage of the sky dome, I never go past 2. Sometimes, even ABE produces a decent result. Ensure your flats are correct and that you are not trying to flatten real signal.

 

[smccully]

Flats are fine, and with LRGB and sometimes Oiii Frames, especially if taken near a Full Moon always require multiple runs of DBE.

 

[smccully]

Here is an example, of a Oiii frame and DBE model most of the gradient removed in the second Oiii file.

Dropbox - File Deleted - Simplify your life

www.dropbox.com

Being able execute a process multiple times, i.e. cycle or iterations is common among many processes in PixInsight.

 

[pfile]

without seeing the control points you used and the tolerance settings in the first pass it's impossible to say if 2 runs were necessary...

rob

 

[pfile]

looking at the original image, you do have problems with the flats... either because your sky conditions are super light polluted and there's lots of stray light entering the telescope during the exposure that the flats can't model, or because there is just something wrong with the flats. the bright corners are not correct. i definitely have problems with LP and stray light and my RGB images look a lot like this one but my narrowband images are generally pretty clean.

anyway try these icons and see what you think... they are on workspace09 at the top left. the 2nd pass didn't do too much, but helped a little.

rob

 

[smccully]

looking at the original image, you do have problems with the flats... either because your sky conditions are super light polluted and there's lots of stray light entering the telescope during the exposure that the flats can't model, or because there is just something wrong with the flats. the bright corners are not correct. i definitely have problems with LP and stray light and my RGB images look a lot like this one but my narrowband images are generally pretty clean.

anyway try these icons and see what you think... they are on workspace09 at the top left. the 2nd pass didn't do too much, but helped a little.

rob

The flats probably should be retaken, though the Moon was pretty close to Full and the target was pretty low in the sky pointing towards the Portland SkyGlow in my area. The model you used is a lot more aggressive then the ones I normally use as well, but does make the point. I still think adding an iterations or cycles to the DBE model would be useful, it may even be more useful to have the DBE iterate until some min. is reached between iterations.

 

[pfile]

ok that probably explains most of it. as mentioned i have the same kind of problems and often have to do very agressive DBEs to clean up my data. someday i'm going to take this telescope apart and flock it carefully (and also probably extend the baffle a little bit as that is said to help with this problem, at least on the GSO RC10.

rob