Author Topic: D.B.E. short question  (Read 2037 times)

Offline gio24

  • Newcomer
  • Posts: 4
    • View Profile
D.B.E. short question
« on: 2014 March 02 02:14:30 »
Hello,
I have a short question: is it better to perform the D.B.E. process on each channel (R,G,B,LUMINANCE) before the RGB combination or to perform the Rgb combination and then D.B.E. on this?
The doubt comes from the fact that if I perform it before RGB the strars color after the RBG combination is too strong while if I perform the D.B.E. directly on RGB the star color is too weak.
Does it give any benefit to perform before A.B.E. with division method and then D.B.E. with subtraction method?
many thanks for your precious help
Giorgio

Offline bitli

  • PTeam Member
  • PixInsight Guru
  • ****
  • Posts: 509
    • View Profile
Re: D.B.E. short question
« Reply #1 on: 2014 March 02 03:33:11 »
DBE works on the R,G and B channels independently, using the same points that you defined  interactivel).  So you can work on an RGB image in one go, and maybe the L separately if it has a different gradient.
Some people are using DBE on the calibrated images, before integration.  The benefit is that it is much easier to handle sky gradients changing with time. The images should be aligned (especially if you use drizzling) so that the background areas you select (using any image) are valid on all images. You can then use the ImageContainer to apply the DBE on the set of calibrated/aligned images in one operation. DBE will recalculate the background values on each image and apply them to the corresponding image (if I understand it correctly). This is relatively fast, thank to the ImageContainer, and give very good results.

I forgott: you should also crop the images to the same area so that they have no black border before applying the DBE. This can also be done with the ImageContainer.

DBE works very well if you have a single kind of gradient, typically the sky gradient (light from the moon, the sun, a nearby city or a combination of this). If you have two kind of gradients, one additive (as the sky gradient) and one multiplicative (due to bad flat fielding), then you are in a search and try mode and all bets are off. Try ABE, DBE, etc... multiple time if it helps.
-- bitli


Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4623
    • View Profile
Re: D.B.E. short question
« Reply #2 on: 2014 March 02 09:56:27 »
i could be wrong about this but i'm pretty sure DBE does not work right with ImageContainer. somehow the DBE process icon is "married" to the image you created it with and the computed gradient is the same every time.

hopefully Juan can comment. i've always had to use ABE if i wanted to do background subtraction on my calibrated subs. and in that case i have not used the aligned subs since ABE seems to be okay with samples falling on stars and such.

rob

Offline bitli

  • PTeam Member
  • PixInsight Guru
  • ****
  • Posts: 509
    • View Profile
Re: D.B.E. short question
« Reply #3 on: 2014 March 02 10:21:15 »
DBE works with ImageContainer.
When you have selected your points, also make sure that you selected "replace the target image" and "discard background model", drop the process instance to the worspace (do not execute it on the image), close the DBE interface, then drag the DBE icon on the ImageContainer.
-- bitli

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4623
    • View Profile
Re: D.B.E. short question
« Reply #4 on: 2014 March 02 10:46:03 »
yes i know about those things.

you don't even need to use imagecontainer to see the problem. do what you described except do not discard the background model; do not execute DBE on the image, just set up the sample points and drag the icon to the desktop.

then open two flats from different OTAs, or any two different images, and drag the process instance to the first image and then drag the process to the second image. use pixel math to divide the background models. you will get a perfectly white image - the two background models are identical. now instead of that, double click the process icon with image 1 selected and apply it from the DBE tool, then cancel DBE. then double click the process icon with image 2 selected and apply it from the DBE tool. now divide those two backgrounds. you'll get some complex image.

for some reason, with DBE, saving the process icon and dragging it to an image does not work. a process icon for DBE can properly save the sample locations but in order for it to have a different result it has to be loaded into the DBE tool and applied rather than simply dragged onto an image.

try it for yourself.

rob

Offline bitli

  • PTeam Member
  • PixInsight Guru
  • ****
  • Posts: 509
    • View Profile
Re: D.B.E. short question
« Reply #5 on: 2014 March 02 11:05:02 »
You are right, I was using a script that recalculated some values, as the values are unfortunately saved in the process instance and there is no way to request a re-calculation from the script.  The script was quite specialized and is not useful in a general context.
Somebody pretty experienced told me about doing the DBE on the individual image, I assumed it was done by the ImageContainer. I need to find out if it was done manually (which is pretty time consuming) or if I did miss something.
-- bitli

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4623
    • View Profile
Re: D.B.E. short question
« Reply #6 on: 2014 March 02 14:46:57 »
it is strange because dynamicCrop process instances seem to work across images.

juan may have explained something about this, i have a hard time remembering things these days

rob

Offline bitli

  • PTeam Member
  • PixInsight Guru
  • ****
  • Posts: 509
    • View Profile
Re: D.B.E. short question
« Reply #7 on: 2014 March 02 22:38:27 »
Quote
i have a hard time remembering things these days
Me too... In fact here was the original discussion with Juan that I forgott

http://pixinsight.com/forum/index.php?topic=6477.msg44042#msg44042

--bitli