Problems With DBE

[pscammp]

Hey all,
    Just become the latest deciple in the Pixinsight gang as of yesturday, even though it's going to take
a while to find my feet i have really enjoyed playing with the trial so it had to be done.

 

So for my first frustration i would like some advice on the DBE tool, i have followed a few tutorials on
this tool but still cant seem to remove the gradients from my M33 image. Im posting a dropbox link
here in the hope some pixinsight master can have a go at it & hopefully guide me to where im going
wrong. The Fits image was produced in Pixinsight from all my subs & calibration files. The result after
the DBE process still shows the red & green'ish gradients in the image although they appear to be less,
no matter what i try it wont remove them any more.

My Canon 1000D had a 2" IDAS LPS-P2 screwed on the front of a Skywatcher Coma Corrector and its
mounted in a Revelation 10" Newt @ F4 (254mm x 1016 FL). Light pollution is a medium issue from
my back garden in case that is important too but not huge.

https://dl.dropboxusercontent.com/u/4938813/light-BINNING_2.fit
Please help someone.

Many thanks
Paul

[pfile]

well i had to do DBE twice but yes the background is still lumpy.

i'm certainly not an expert with DBE, there are a lot of knobs and it's never 100% clear to me what to do.

in the end after the 2nd DBE i did color calibration and processed the image. at the end i used atrous to remove the color casts from the background. this removed some of the color from the galaxy despite a very strong mask. i'm sure with range mask i could have done better but you have to put the stars back in the mask or you'll remove the colors from the stars as well. even so i've kind of killed the star colors.

so not my best work but there is a way forward with this type of image.

other things: the image seems typical of a DSLR image taken thru an LP filter. the red channel is very weak/noisy. if you can increase your sub exposure times that will help a lot. also the shadow of the mirror is falling on the bottom of the frame but this is typical with fast systems with wide FOV.

rob

[Harry page]

Hi

First off crop the image to remove edge defects and then

1) Apply DBE increasing tolerance and lowering the smoothing factor the enable better modelling of those rapid changing colour problems  also lots of samples req in this case ( see below )

2) split into RGB and use Linear fit to balance the channels to remove colour bias ( see this vid http://harrysastroshed.com/RGB%20Combination.html ) 

3) Recombine into RGB and process away to your hearts content

Regards

Harry

[pfile]

harry i have not looked at your xpsm yet but i find that even with a lot of samples and a lowered smoothing factor that there's still un-evenness to the background. i faced that on my own M33 image which was shot from heavy LP.

rob

[Harry page]

Hi

I think the image as attached has as good as a background as can be expected from the data

I could quite easily work a decent image from this , its always tempting to push to hard 

Harry

[pfile]

yeah okay, that's about where i was with it.

rob

[pscammp]

Thanks Guys,
    Been playing the hell out of the DBE this last week and it's amazing what you can do
with a little advice from the experts and lots of time applying & reapplying different settings.

Finally i have a result i feel happy with and i even managed to get rid of 95% of the dark bar
which was evident a little above the bottom of the image. I feel i have a reasonable result under
circumstances of what i was dealing with in the first place. As well as auto generating samples at
25 rows, i also added over 100 new ones in the corners and on the bright bar at the bottom of the
image. I also added loads of samples over the dark horizontal bar too which seem'd to do the trick
in the end. Overkill ? probibly but there you go.

Here's a fits file after applying Image Crop, DBE, BN & Color Correction. The file was saved as a
32bit float fits file with linked STF applied.

https://dl.dropboxusercontent.com/u/4938813/light_BINNING_2_DBE.fit

So have i got myself in a 'newbe induced false sense of security' or did i manage to get a reasonable
result 

Many Thanks everyone
Paul
 

[pfile]

i think that looks pretty darn good. the red channel is a little noisy so it's hard to eliminate all the background gradients...

the dark bar is probably caused by the shadow of the (retracted) mirror. for a normal lens it's out of the way but for some telescopes the light cone impinges on the mirror.

rob

[pscammp]

pfile,
    This 'Dark Bar' a bit of a pain in the 'As*'. I have an M31 image integrated etc in PI and although i can 'just' see this bar it's so faint
that you'd miss it if you were not looking for it.

How can a flipped up mirror cause a thin dark shadow part way up an image when the flip mirror is all the way up the top & out of the
way 

Is there anything ref the DSLR that can be done to solve this problem at the hardware end or is this something we have to live with.

Cheers
Paul

[pfile]

it's not totally out of the way; there is some angle at which light will hit the mirror.

some people have gone as far as to remove the mirror by cutting it out of the mirror box! the camera is apparently okay with that.

anyway maybe i misinterpreted your dark bar; it should lie right across the bottom or top of the frame. it might start 'migrating' if your subs are not well aligned though. meaning, the center coordinates of the subs differ a lot.

rob