Back to forums

M33 and StarXterminator

Processing Deep Sky M33 NGC598 Messier Objects
14 replies•619 views
Blaine Gibby avatar
Blaine Gibby avatar
Hello, I am starting acquisition on M33 and my goal is to obtain Ha, O and LRGB data to combine. It will be important to remove the stars, however StarXterminator removes many important features of the galaxy. Has anyone had luck removing stars while processing M33? Also BXT seems to be pretty heavy handed on the galaxy details. 

For example here is the stars generated by StarXterminator: 
Well Written Respectful Engaging
Francesco Meschia avatar
Francesco Meschia avatar
M33 is a real challenge for star-removing NN, because we can actually see individual bright stars in the galaxy. But I did have good results by importing in Photoshop the stars and starless (stretched) versions, setting up the layers to show me the differences between the two, and painting a third layer to cover the portions that I believed were features of the galaxy and not foreground stars. Then I imported this layer in PI, converted it to grayscale, and used it as a mask to prevent SXT from removing those features.
Of course it’s highly subjective. And boring, and tedious. But it works. See my M33 at https://www.astrobin.com/0hpqzd/O/
Well Written Helpful Insightful Engaging
Jared Willson avatar
Jared Willson avatar
Yeah, with this galaxy I have noticed a few things:
  • StarXTerminator is not as useful as with most subjects. There are simply too many "stars" that are right on the edge of being stars, or being small bits of detail in the galaxy. I would recommend not actually using SXT and just stretch and process the stars and nebulosity together. Note, this does not necessarily apply to H-alpha and OIII data--depends on how you do continuum subtraction before blending back into the LRGB data
  • BlurXTerminator is a disaster on M33 in terms of "worm" creation if you use it for both star correction and enhancing non-stellar detail. I find I need to drop the amount to 0.1 or so on non-stellar details or BXT will try to link up what are probably moderately bright stars right on the edge of being resolved, and this results in an oatmeal like background to the galaxy.
  • BXT does better (introduces fewer artifacts) if you are just a tiny bit heavy handed with NoiseXTerminator when you are still in the linear state. I would normally apply NXT at perhaps 0.3 while still linear in order to avoid artifacts with BXT afterwards, but I find I need a value more like 0.5 on NXT to keep the artifacts from appearing in the BXT step.
Helpful Insightful Engaging
Jeff Reitzel avatar
Jeff Reitzel avatar
Hi Blaine,
 Good to see a local person here. An option is using a mask to protect the galaxy from star removal. I personally use the GAME script for circular or elliptical masks on projects like this. http://skypixels.at/pixinsight_scripts.html    I know there are a few other options out there for this purpose as well.
As far as BXT goes you can turn down the settings a bit to make it less aggressive. For galaxies it seems to perform better if you manually enter the PSF as opposed to auto PSF. At the same repository is a script called PSF Image Creator that will quickly calculate the average PSF of your image. Or you can do it manually using some Moffat stars close to the galaxy for best results. I'm sure you saw my last image of M33 at our local website using the manual method and masking for an example.

CS,
Jeff
Helpful
Rostokko avatar
Rostokko avatar
I have found StarNet2 doing a much better job than SXT on M33.
andrea tasselli avatar
andrea tasselli avatar
I have found StarNet2 doing a much better job than SXT on M33.

So very true.
Tony Gondola avatar
Tony Gondola avatar
I've been working on this object for the last few days and am finding that not doing star extraction is the way to go. There are so many borderline stellar objects in the galaxy, other than gross area masking, I don't know how any method is going to discriminate correctly. M33 isn't in a particularly star rich region so this seems fine to me. You don't always have to extract the stars IMO.  What I am struggling with is getting my mosaic panels to blend!
Helpful Insightful Respectful Concise Engaging
Oskari Nikkinen avatar
Oskari Nikkinen avatar
You can kind of hack BXT to work with M33 and avoid the worms. Run BXT first in correct only and then measure FWHM with the FWHM eccentricity script. Then run BXT with automatic PSF unticked and set it manually to somewhere around 50-70% of the previously measured value. The worms are either gone or much better, at least on the last image i tried it on (not posted this one yet).

For SXT i dont really have a great solution. With my 8" or 10" newtonians almost all of the interesting detail is in the star layer, so the starless layer only sees a minor touchup with HDRMT to increase dust lane contrast. Not sure how i'd go about adding the narrowband images to the broadband data. Probably with Photoshop, but should say i am not too experienced in this kind of work.
Helpful
Blaine Gibby avatar
Blaine Gibby avatar
Thanks for the insight everyone. Sounds like I have some innovating ahead of me smile

Hi Jeff good to see you on here as well. Great job on your Sharpless planetary nebula! I’m in Filer, just west of Twin Falls
gfunkernator avatar
gfunkernator avatar
You can perform a star removal early in the process like before any level stretch or give it a very subtle stretch.  You play with the levels so that only stars in our galaxy are visible enough to be considered by the tool hoping it won't touch M33.  I would try it myself but my M33 stack is pure poo.  What I have posted on my profile is a lot of post magic.
Brian Puhl avatar
Brian Puhl avatar
Try not using star removal in your process.   or IF you do decided to use star removal on galaxy data, do it after the image is stretched.    There are tons of background galaxies in  your image and 80% of them get eaten by star removal.   Once they're in your star layer, anything you do to that star layer also affects those background galaxies.  If you extract stars linearly, then stretching your stars softer (which is something most people do) will also mute the luminance of the background galaxies.  By the time blurX gets in the mix, then reduction, contrast, etc.... all those small disc shaped galaxies turn into round stars.  


Had I used starless on this image (work in progress) those small background galaxies probably wouldn't be there.    The depth of the galaxy will likely be muted since there will be a plateau of sorts where stars exist and then drop off after losing them due to denoise and such on the starless layer.




Even if you don't care about these things I mentioned, I high recommend you at least TRY not doing a starless process on galaxy data.   You might be surprised.
Helpful Insightful Engaging Supportive
Tony Gondola avatar
Tony Gondola avatar
I think it's important that star removal not become so automatic in our workflows that we do it without thinking. Everything should be done for a reason and every subject is different.
Well Written Insightful Respectful
Harel Boren avatar
Harel Boren avatar
Tony Gondola:
I've been working on this object for the last few days and am finding that not doing star extraction is the way to go. There are so many borderline stellar objects in the galaxy, other than gross area masking, I don't know how any method is going to discriminate correctly. M33 isn't in a particularly star rich region so this seems fine to me. You don't always have to extract the stars IMO.  What I am struggling with is getting my mosaic panels to blend!



I have also run through the processing of M33 without star removal. Here’s my result:
https://astrob.in/09ihp9/0/
Tony Gondola avatar
Tony Gondola avatar
And mine:
https://www.astrobin.com/bn5f7t/

I gave up on the mosaic but I like this view of the core. I really don't think that in this case, star removal would have helped the image at all.
NeilM avatar
NeilM avatar
I processed M33 recently in Siril and I used StarNet++, which I was happy with.

https://www.astrobin.com/axsbsb/B/