The Antennae (NGC 4038, NGC 4039)

[marc4darkskies]

Despite a never ending series of poor seeing sessions since March, I managed to get enough (mediocre) data to finish this off. I think it's a respectable result for a relatively small object, but it was a battle - I used 39 subs and threw away 41 subs that were worse than 3 arcsec FWHM).

Enough complaining - here 'tis ...

Full res, cropped for composition:
http://www.pbase.com/gailmarc/image/163198418/original

Cheers, Marcus

[TR]

What a luxury, to be able to throw data away ;) It is a small object, nice result.

Terry

[RickS]

Very respectable, Marcus! Some lovely detail in the cores.

[Atmos]

:thumbsup: Has turned out a treat Marcus, nicely resolved and good colour.

Out of curiosity, what it did look like if you ran more than 50% of the subs together? I suppose it largely depends on what the FWHM of the subs you kept were :)

[IanP]

It's a very respectable result, Marcus. Great shot :thumbsup:

[marc4darkskies]

Thanks Terry. Not a luxury - a massively frustrating and extremely annoying waste of time. :sadeyes: Onwards and upwards though!



Cheers Rick!



Thanks a lot Colin! Adding any of the rejected subs (3 to 5 arcsec FWHM) simply reduced detail and contrast with faint stars and background galaxies. As it was, it was a struggle to show what you see in this image.



Cheers Ian!

[Peter Ward]

That's might 'purdy there partner.

Very slick indeed :thumbsup::thumbsup:

[gregbradley]

The galaxy itself looks pretty good. My CDK is also quite sensitive to seeing so its just one of the limitations of this type of imaging short of putting it on a mountain. Narrowband though seems to cut through the seeing quite well.

Greg.

[Placidus]

Crispy sharp in the core, deep feathery antennae. Everything one could want. Beautiful colour too.

[Paul Haese]

The detail and colour in the core is very nice Marcus. The rest is just more data as you know. Still good given the struggle.

[strongmanmike]

Well, this image sounds like it has been a bit of a trial for you Marcus? :scared: :bashcomp:...some nice details showing in the central areas none the less :thumbsup:

Mike

[topheart]

Good job Marc, especially considering the poor seeing .....hope Sydney gets better seeing where you are soon.
Cheers,
Tim

[marc4darkskies]

Thanks Peter!



Cheers Greg. A mountain would be nice but I'd settle for 900m or so!:)



Thankyou M&T! Glad you like it!



Thanks Paul.



Cheers Mike! Don't get me wrong, I'm happy with it but I know I can do better if the conditions cooperate. You make the best with what you're given though, right?



Thankyou Tim!

[rustigsmed]

very nice work Marcus, love the resolution in the core. annoying about the subs!

[codemonkey]

Very nice, Marcus! Nice colour and excellent detail in the galaxies.

The antennae look pushed a bit hard for my tastes, the transition to the background seems too abrupt when looking at the full res.

I recently found good use of bad subs by integrating two sets of data: the good subs, and all the subs. I then made a mask out of a stretched copy of the "all subs" integration (which had bigger stars, blurrier detail), and then replaced the higher signal areas of the image with the good set.

Granted, it doesn't solve all the world's problems, but it did mean that I got a much cleaner background, while preserving detail in high signal areas. Good to smooth out things like the antennae as well.

[marc4darkskies]

Thanks Russell!



Thanks Lee.



Fair enough. It can be a difficult balance to get the transition of such faint elements to be as smooth as they are in reality without megadata. I opted for showing as much as I can with a mere 9 hours of data, even showing faint stars blurred by bad seeing. It can also be a function of the monitor your using too but I'm assuming you have a calibrated monitor. You can check contrast & brightness settings here (http://www.pbase.com/gailmarc/image/75937224):



Blending bad subs with good subs is fraught - especially when it comes to the faint background stars and galaxies. Isolating them from the blurring effects of bad subs very problematic.

[Stevec35]

More than acceptable Marcus - I think it's very nice. I hear you about bad seeing. It's generally the norm in my part of the world.

Cheers

Steve

[batema]

I think it looks very nice. If you could outline which program you stack your data or how you determine each frames fwhm data that would be great. I know its basic but I use Deep Sky Stacker and it gives me a fwhm value but mine are just huge. I don't think I have seen below 3 or 4 sadly. I also capture using MaximDl. Thanks.

[DaveNZ]

Good work Marcus. Nice colour and good detail even though seeing was average.

[marc4darkskies]

Thanks a lot Steve! I usually get a smattering of good seeing nights, but not lately. Hmmm, I would have thought Canberra would have better seeing conditions on average than here.



Cheers Mark! I use CCDStack for calibration, stacking and initial deconvolution. It shows FWHM in pixels. For a more reliable measure I use CCDInspector which measures FWHM in arcseconds (with appripriate parameters describing your optical configuration).



Thanks very much Dave!

[Ross G]

A beautiful photo Marcus.

Great colours and some amazing detail, especially in the core.

Even more amazing considering the pathetic Sydney skies we have had this year.


Ross.

[marc4darkskies]

Thanks very much Ross! I think I need to move to better skies sooner rather than later!

[Stevec35]

[QUOTE=marc4darkskies;1256821]Thanks a lot Steve! I usually get a smattering of good seeing nights, but not lately. Hmmm, I would have thought Canberra would have better seeing conditions on average than here.

Canberra is hilly, hence plenty of turbulent air and I'm on the side of a hill which is particularly bad.

[strongmanmike]

[QUOTE=Stevec35;1257291]

I can say that the incidence of decent to good seeing at 600m ASL in the Wallaroo area north of Canberra and when I was at 750m ASL in the Googong area south of Queanbeyan in 2004-2006, the seeing is significantly better much more often than at SL in suburban Newcastle, where it was crap all the time essentially.

So moral to the story...you need to site test before deciding on where to build your observatory :question:....:lol:

Ok, not feasible :sadeyes:

Mike

[Peter Ward]

Definitely not coming the raw-prawn with this one :D

A subtle yet deft improvement. :thumbsup: :thumbsup:

[alpal]

I agree - I think Marcus could be tempted to have another go at the processing &
not throw out all those subs.
There is still faint luminance information in subs with poor FWHM.
Still it's a pretty picture considering the seeing problems.

cheers
Allan

[marc4darkskies]

I did indeed make a subtle change last week :D. Amazed / impressed that you'd notice the difference! :thumbsup:



Thanks Allan. No, I won't be tempted. :lol: I'm an empiricist. That is, I don't make purely theoretical decisions :einstein: :nerd: to negate possible courses of action. I'll suck it and see to prove it doesn't work :bashcomp:. Adding very poor seeing subs (where faint stars - that were barely above background noise - are essentially blurred to oblivion) to bring out faint extents or background does not / did not work. I.e. there are no free lunches.

[codemonkey]

Low signal detail can be lost just as well through poor SNR as well as through using convoluted data.

End of the day, it's a trade off, and maybe you're not happy with making it, which is fine, some things matter more to some people than others.

The success of this method will largely be based off how good you are at creating masks to preserve the areas you care about. Remember that you have total control here, so you could literally protect every detail if you're fastidious enough.

Anyway, it was just an idea to consider, a potential to scrape some benefit out of those house of less than great data.

[gregbradley]

I'm with Marcus on this. I have done the same as he has and many times. Versions with all the data of which some are poor and some with only the better. I find invariably the ones with only the good data are better. The only cost usually is extra noise. I think overall you are better off wasting some data (obviously within reason) and reshooting if needed to get the required data to get the low noise image. What that amounts to depends on the system, speed of the scope, aperture and sensitivity of the camera. Also the seeing. I definitely prefer less high quality data than a mix of good and poor. Rick S though went over how in PI you can weight the mix so that may salvage a bit more if you use FWHM values to weigh the mix of data.


It would make a good thread-what does it take to get the ideal image?

Greg.

[alpal]

The faint glow of the outer extremities of a halo has no detail in it anyway.
Surely by stacking - more of that faint halo could be revealed?

[marc4darkskies]

Perhaps I didn't make myself clear Lee, none of the data I used was "great" data - none of it. There were a number of subs I used that were well outside my normal tolerance for acceptability. If you look closely you'll see that manifest in relatively poor depth and quality of faint stars. Don't worry though, I've teased every morsel of depth and detail out of the available data without rendering artefacts or making it look like I've tried very hard - maybe that's the problem :lol:. Finally, complex masks and layer blending are child's play in PS and the norm in my processing, but there is a limit to the overall magic you can conjure using bad data without making an image look forced in some way.



Thanks Greg! :) Time or the paucity of clear stable nights will tend to limit ones ability to gather megadata, so honing noise mitigation techniques is essential to good processing technique.



There is in fact large and small scale structure in the antennae. E.G. some of the knots you see on the outer fringes are actually blurred stars that, with good data, would look much clearer. I try not to erase these, even though they may look like noise.

Final thought is that people, especially beginners, need to understand that while good processing is paramount to producing a good image, no amount of sophisticated processing will overcome the limitations imposed by bad data.

[alpal]

Dear Marcus -
who said anything about losing or erasing stars that are there now in your picture?
You can stack all the frames & even blur the image you create but
use it as a blurred layer mask to increase the intensity of which ever areas are whiter - in that mask by clicking luminosity in Photoshop for that layer
& then increasing the brightness using curves.
This will bring up the faint halo better for the tidal streams.
It has to work.
By the way - it's not cheating or devaluing your sharpness of the picture you already have.
It's using the so called lost information to increase the brightness in the areas you want.

cheers
Allan

[marc4darkskies]

Hmmm ... you're not understanding the processing challenge Allan. You're giving me a beginners lesson on how to stretch an image using masks and curves because, presumably (?), you think there is more faint extent to reveal in my data and that this can be achieved using bad data. Firstly, the image already compares favourably with others on the net in regards to depth so, frankly, I don't actually feel the need to stretch any more, especially with this mediocre dataset. Moreover, I'm telling you that I can't use the bad data to improve my image (noise levels mostly) without losing faint stars, galaxies or even structure in the streams.

The subs I excluded ranged from 3 to 5 arcsec FWHM with most somewhere in the middle. The sky was also somewhat brighter during most of the bad subs - quite common. Consequently, S/N is much worse and the faintest stars in the stacked bad data are dead & gone. They have passed on. They are no more. They have ceased to be. They are at best fuzz balls that are hidden at or near the sky background. To illustrate, see the two screen captures below of the stacked good and bad data (autoscaled in CCDStack with, no other processing). The capture is from the tip of the right hand tidal stream where the reeeeally faint lobe is. No prizes for guessing which is the crap data (31 subs worth compared to only 16 good data subs).

Now, you want me to layer in the stretched bad data (ie blurred with poor S/N) where faint stars and structure are not visible, to achieve some kind of miraculous recovery of depth/noise without losing faint star detail. Theoretically I can do that layering, but not without painting into the mask those stars and structures that are very faint so I don't blur them out of existence. Note that no automated mask creation can pick out ALL of the faintest stars & structures I'm talking about unless your S/N is much better than this example (ie: would need much more data).

On top of that, I'd have to deal with this challenge over the entire frame. Even if I could brighten the faint stream extents I would drown out the good faint stars unless I brighten them too. I'd do that by inverting the mask and applying it to a curves layer and brightening every other star in the field - unless I mask out the rest of the field as well. It's quite probable that I'd also add signal into the streams (and elsewhere) that isn't real because the faint extents are practically invisible above the sky brightness in the bad data.

Now, if you look at that region in my processed image, you'll see the lobe quite clearly (albeit with noise) AND the faintest stars in the screen shots. I'd be chasing my tail for no benefit and probably end up still losing faint detail and/or adding detail that's not real.

[alpal]

Dear Marcus,
I've been having trouble with internet connections.
Both my Optus cable broadband & my backup Vodafone stick failed to work.
I'm back on now with a new SIM card in the Vodafone stick.

I was just making a suggestion that I thought was worth trying.
It's your picture so you can decide for yourself what you want to do.
My humble opinions are only that.

cheers
Allan