The Andromeda Galaxy – Oct DSC

[JonH]

This is not the final image, I am just after everyone's opinion about the colour balance and such as without a calibrated monitor I pretty much guessing!
There is still a lot of work to be done with this and i think there is still a fair amount of data to be brought out, but it shows that I have captured detail almost right to the core which is exactly what I wanted!
Had I not lost the 10sec subs the core might be better!

90x300sec
20x150sec
16x30sec
All stacked and processed in Iris with a little final tweaking in photoshop.
The different length subs were stacked and combined in Iris using the desc_hdr merge_hdr and reduce_hdr1 commands. Not very easy to figure out but when you get it right it does a pretty good job! I still think a full re-stack might be in order to really utilise these commands though.



Full size (this is still a heavily compressed JPEG, just bigger!):
http://farm9.staticflickr.com/8473/8097434821_87fec7a005_k.jpg

[MarkS]

Looking pretty good!!

I would say the colour is more or less right - just some minor tweaking required.  You've coped well with the huge dynamic range.  The core brightness is fairly well controlled and you have good star colours.  But you have a strange artifact around the edge of the galaxy - some kind of posterisation - visible steps in colour and brightness - you should check which stage in the processing introduced this.

Looking forward to the next version ...

Mark

[JonH]

Yes I noticed posterization, it appeared funnily enough while trying to tame the blue. Even before I had started to stretch the colours I had an abundance of blue stars, and no matter what I tried it only got worse.
I eventually managed to get it under control in the stars, but it created that as a side effect. Is very noticeable in the blue along the right!
Also in the smaller M110 there is something very odd going on where it appears all the colour to the central part has been lost then suddenly appears again at a very defined point.
I'm not 100% sure when this happened as I was paying so much attention to getting everything right but I have an idea why it happened.

[Carole]

That's pretty amazing considering how long ago you started imaging.  The core looks very good to me, and the detail in the galaxy, I think you've tried to darken the sky colour too much and I wonder whether this is contributing to the artefact problem. 

Excellent work.

Carole

[Mike]

Very nice Jon and especially for a newcomer.

Please don't forget to put your details re: Scope, Camera, Mount, etc.

Thanks. Mike

[JonH]

Thanks

Yes good point:
Equinox ED80 with 0.85 flattener/reducer (spacing yet to be set correctly)
NEQ6 with eqmod
Modified Canon 450D
Guiding with QHY5 on ST80
Captured with APT

[JohnP]

Blimey Jon,

That is excellent even more so seeing how short you have been imaging. Like Mark says a few weird things going on around edges (look very purple). The focus & alignment is excellent. To compare this was my first ever attempt at it...

http://jpastronomy.co.uk/graphics/archive%20images/dso/galaxy/m31_007.html

Keep up the good start & you'll be giving Mark a run for his money in no time.

John.

[MarkS]

I forgot to mention that your star shapes are pretty good right out to the corners.  There are no obvious distortions.  This tells me your reducer/flattener is pretty well matched to the optics of the Equinox ED80.  Especially since a DSLR has such a large chip size.

[JonH]

Thanks John, it's getting there!

Keep up the good start & you'll be giving Mark a run for his money in no time.

I dunno about that, the difference between mine and Mark's images even before the processing is finished is huge!
That said I'm pretty happy with the results I'm getting so far.

I'm very happy with the star shapes in this, it's the best I have had to date by a long shot!
What's weird is the stars are better than in any individual sub?
I was wondering if by using the quadratic global matching in Iris with so much data it has actually 'fixed' them for me?

[MarkS]

What's weird is the stars are better than in any individual sub?
I was wondering if by using the quadratic global matching in Iris with so much data it has actually 'fixed' them for me?

It's not weird at all. Star shapes in a single sub are messed up by the CFA (colour filter array or Bayer matrix) over the sensor's pixels and the subsequent pixel interpolation that is performed during the construction of the colour image from the raw data.  When many subs are added together, especially when dithering was used during acquisition, these effects are averaged out.  It's one of the important reasons we use dithering

[Ivor]

Good image Jon, I'm looking forward to the final result with the suggested changes above..

[RobertM]

Great image Jon, you're really coming on leaps and bounds

To be critical I would say that the background is a bit clipped and because of that you've lost a lot of background data but the rest is looking really good.  I also think HDR should give more core detail than you have.  I've found M31 to be a bit of a pig to colour balance in the past but that's a pretty good version as seen with this crumby monitor.  If you can sort the background out then the rest should be much easier.

Robert

[The Thing]

Good job Jon!

I haven't got round to my measly 9x300s subs yet and looking at this I am bound to be disappointed with the result..

[JonH]

Yes I agree about the background, it was done deliberately as I was loosing patients trying to get rid of the gradients. In the end I just gave up and darkened it until they couldn't be seen anymore.
In hindsight this was a stupid move as lost a lot of fainter stars and fainter detail to the outer edges.

[JonH]

Thanks Duncan
To be honest you unless you're up for a huge challenge, spending hours on end processing and even more time waiting for your computer to register and stack, 9 subs is the way to go!
You can only work with what you've got...

[JonH]

I am slowly coming to some conclusions here, either Iris is not cut out for HDR images or I lack the ability to get it to!
To be honest it is more likely the latter 

Either way no matter what I try I only get so far before hitting a dead end! 
Think I might have to scrap the merge_hdr idea as it just isn't giving the results it should, for me at least!

[MarkS]

Iris is not really cut out for HDR images.  It's native representation is only 16bit integers/channel.  I've never really played with HDR in Iris because the first thing it forces you to do is to downsample back down to 16bit again.

[JonH]

Well at least I know it's not all me then...
I suppose that means I need to stick to using Iris for more basic processing and stacking then, and find a new bit of software to learn!
Any suggestions? The free'er the better!

[JonH]

And here we have process number two!
Yes I know the colour balance it miles out, but for this one I concentrated more and finding all the detail there was rather than messing about for ages trying to get the colours just right....
There is still much work to be done, but I have some really good ideas how I can improve this further, number one being to take more shorter subs and across of a wider range of exposures.
Either way I think I finally might have found a way to create a half decent HDR image! 



Larger version:
http://farm9.staticflickr.com/8047/8101210417_b19cefb6b3_k.jpg

[MarkS]

That's beginning to look very good! The colours within the galaxy and the star colours are great and it really is a stunning image for someone who has just started astrophotography.

It also clearly demonstrates the value of taking hours and hours of exposures in the battle against light pollution.  

Well done.  You ought to be really proud of it.

Mark

P.S. now you've just got to sort out that posterisation

[JonH]

Thanks Mark, It's finally getting somewhere close to what I am after!
I found there was a way of getting photoshop to create a HDR image at 32bit so until I get more powerful software I think that is the way to go.
This was only my first try at doing it this way and I think with some very careful manipulation I can get a touch more detail in the core, but I'm pretty sure I have found the outer extents now.
I'm really not liking uploading the jpeg's as it just trashes the data! I need to get me some server space somewhere to upload full size Tiff's, this one was over 140mb!
But if you could see the original those three little stars right in the core are visible!

I also want to get MORE exposure time on this! 

[Fay]

you really have done well Jon, background a little too dark and second version better than the first, great M31 even if you had been doing it for a few years!!!!!

[MarkS]

Photoshop is a good idea for HDR.

If you create a blurred version as a new layer e.g. a 5 pixel Gaussian blur and subtract say 50% of it from the original you will increase contrast in the dark lanes.  Try it in the brighter core areas but you might have to first mask the stars to prevent dark rings appearing round them after subtraction.  Don't sjk me for more details because I'm not a PS user.

Please don't bother uploading a 140Mb file - no-one will ever download it!  Just reduce the amount of compression you use on your JPG.

[JonH]

I might have to try this, the problem with PS though is although it will create a 32bit image from my three stacks it will let you do very little with it!
Although i do wonder if 32bit is enough to capture the full range? If i have three stacks each at 16bit then surely the final image would have to be 48bit to truly preserve everything?

[JonH]

I'd like to say this was process number three, but it's not, perhaps 300! But it is one that i am happy enough to post up.
86 x 5min subs stacked in Iris then processed in ImageJ with some final adjusting in Photoshop.

You might be interested to know that i haven't used any shorter subs in this at all, only the 5min ones!
I have spent a lot of time exploring out how to compress the dynamic range and still preserve the fainter detail and it looks like it is finally starting to pay off.

I'm not really sure if this is better or worse that my other tries with regard to colour, but it's making progress, i think....



Bigger one:
http://farm9.staticflickr.com/8057/8168177469_6f3696fea7_k.jpg

[Carole]

That's pretty stunning Jon, people who have been imaging for a few years would be well chuffed with that.  I prefer your version of 18th October as there is better colour.

Carole

[MarkS]

It's looking much better now!

[JonH]

Thanks, I'm pretty sure I managed to find all the usefull detail I can but Carole is right, more work it still needed on the colours.
I'm starting to wonder if I'll ever get the right! 

[Fay]

I like bits of the first one & bits of the second. bit more lane detail & contrast in the first, but both very nice

[RobertM]

Looking good but I agree with Fay.  You've lost contrast and perhaps pushed the processing a little far (though I can see what you were aiming for).  There is also now a very noticable gradient  from top left to bottom right.

As you can probably see, this processing malarky can just go on and on as there is nearly always something that can be improved

Robert

[JonH]

I think the contrast can be recovered using some of the HDR functions in ImageJ or with the method Mark sugested.
As for the gradient I'm not sure where to start with that.
I'm going to try improving the colour now, unfortunatly the method I have been using to compress the dynamic range has a side effect of killing the colour and it's an ass to get back!
But I still prefer this one as it shows much more of the outlying detail.