Dual-scope image of M33

[Adreneline]

Hi Everyone,

I am experimenting with my new dual scope setup trying to decide on the best way to image with the combination of SW-ED80/WO-ZS71/414osc/428mono.

This is M33 with 21x300s from the ED80+414osc and 20x180s L + 5x600s Ha at which point clouds stopped play.

RGB, L and Ha were processed separately in PI and then combined in PS with a luminosity L layer and an Ha lighten applied to R - I think/hope as per @ollypenrice advice.

As ever any comments on any aspect of the scope setup, process or image would be much appreciated.

Thank you for looking.

Adrian

[carastro]

Great result Adrian.  Nice and smooth.

Carole 

[Adreneline]

19 minutes ago, carastro said:

Great result Adrian.  Nice and smooth.

Carole 

Thank you Carole. The alignment is still not perfect but there is now 100% overlap so I get the full benefit from both scopes. I also moved the filters down adjacent to the 428 and the difference on the lights is significant. A few more Ha frames would have been helpful  - which would also have given a few more rgb frames at the same time. Two for the price of one!

Thank you again for your kind words.

Adrian

[ollypenrice]

That looks as if it's working well but for me the noise reduction is a little intrusive. I've think you've maybe given it more than it needed. M33 is a rum target because it has a grainy surface which looks like noise but isn't. (I say this because it's popular with our guests and when I combine endless hours of data from different shoots it remains grainy-looking.)

I'd say your focus was nailed so don't blur it too much with NR. It doesn't need it. Just my take on this good image.

Olly

[Adreneline]

10 minutes ago, ollypenrice said:

That looks as if it's working well but for me the noise reduction is a little intrusive. I've think you've maybe given it more than it needed. M33 is a rum target because it has a grainy surface which looks like noise but isn't. (I say this because it's popular with our guests and when I combine endless hours of data from different shoots it remains grainy-looking.)

I'd say your focus was nailed so don't blur it too much with NR. It doesn't need it. Just my take on this good image.

Olly

Thank you for the feedback Olly.  I will give it another go although I may hold off until I can get a few more Ha frames to add to the mix; I had hoped to get 10-12 but had to stop after five and even then the last one was spoilt by cloud so there are only four in the stack.

Knowing when to stop with a given process is still something I find difficult. I have learnt to stop a lot sooner with colour saturation and even 'back off' on occasions when I have judged there is too much colour. Noise control/reduction is still a bit of a dark art to me although I do aim for the 23/23/23 background. I think that for this target that is where I've gone wrong maybe?

I have to admit I found it easier to combine the RGB/L/Ha in PS than in PI; it's much easier using a 'slider control' and seeing the impact of the change immediately than using several iterations of PixelMaths.

Thank you again for the feedback and advice.

Adrian

[gorann]

That dual rig seems very promising and time-saving. I have been thinking about something similar. When it comes to processing M33, you may want to have a look at this thread for inspiration:

 

[Adreneline]

Thank you Goran @gorann. I'll take a look at the M33 processing examples; personally I tend to favour more subtle colours in my images but I fear based on advice from Olly I've overdone the noise reduction and so probably lost some structure detail.

Well I am hoping to have more fulfilling imaging sessions by using the dual setup. As I've not transitioned to NB I am hoping to collect just L and Ha with the ZS71+428 combination and lots of RGB with the ED80+414osc combination at the same time. The unpredictable and 'un-forecastable' weather here in this part of the UK is a real frustration at times; the central UK always seems to get the worst of what ever weather is on offer for the surrounding NSEW regions.

The overlap (once I've got the alignment fully sorted) seems good and despite the fact I have no flattener on the ZS71 it seems to be performing very well with the size of sensor in the 428.

The intention is to get a reducer/flattener at some stage so I can use the ZS71 with my Canon 70 and (a) do wider field and (b) have something a little more portable.

Always good to have a plan!

Thank you for looking.

Adrian

[ollypenrice]

The thing about noise reduction is to apply it only where you need it. This is sometimes called the 'zone system.' One easy way is to make a copy layer in Photoshop and give the bottom layer a dose of NR such that some grain is still left in the darkest parts. Then go to the top layer and use the colour select tool to pick the noisiest parts (which will usually be the darkest parts.) Expand and feather this selection slightly and then use CtrlH to hide the 'selection ants' before running the eraser over the top layer at, say 50% opacity. How does it look? For more NR run the eraser again.

You will never want to run NR on areas of strong signal. In fact you may want to do the opposite, which is sharpen them - but be sure to exclude the stars if you do so.

Olly

[Adreneline]

Thank you again Olly for your help and advice.

Since buying PI in January I've used it exclusively for noise reduction applied to my osc RGB lights as guided by the procedure given in the LightVortex tutorials. I wonder if now I've started combining fully processed RGB from an osc with L and Ha I am being heavy-handed with noise reduction in the latter. Should I apply any, or only minimal noise reduction to L and Ha or just go for contrast enhancement using LHE and maybe deconvolution to sharpen the detail? Can PS be used in the same way to remove noise from L and Ha?

Rhetorical question? Do you ever reach the point where you spend less time processing than collecting the data in the first place?!

I suppose that is why we should be grateful for cloudy nights.

Adrian

[wimvb]

On 2017-10-13 at 00:31, Adreneline said:

I have to admit I found it easier to combine the RGB/L/Ha in PS than in PI; it's much easier using a 'slider control' and seeing the impact of the change immediately than using several iterations of PixelMaths.

Have you tried the nbrgb script in pi? It's the same method as described by Kayron Mercieca in his lightvortexastronomy tutorial. Only, with sliders. 

As for noise reduction, I find myself using less of it with rgb mono imaging, than earlier with dslr (osc) imaging. But then, my camera was old and noisy.

My standard workflow included tgv denoise on L only in Lab mode, and mmt on chrominance with an inverted luminance mask, thus using the zone system that Olly mentioned. Both processes on the linear image just before stretching.

[Adreneline]

7 hours ago, wimvb said:

Have you tried the nbrgb script in pi?

Hi Wim.

No, I've not tried the NBRGB script but I will give it a go and see how I get on. I use masking/zoning extensively in PI to try to ensure that NR or colour enhancement is properly targetted. In both cases I feel it is the level of NR or colour enhancement that I struggle to get right, especially NR.

For me PI and PS both have a place in the processing of images in just the same way that combination drug medication can be more effective than relying on only one drug. In both cases it comes down to knowing when to use it, where to use it and how much to use - and knowing when to stop using it!

I love the challenge of solving the problem and I really value and appreciate all the help that members of this forum give so willingly.

Hopefully at some stage I will feel I am getting it right otherwise what is the point of doing it in the first place.

Thank you for your help and advice.

Adrian

[ollypenrice]

Sorry, I was about to ask Wim a question then the answer dawned on me! Please ignore...

lly

[wimvb]

43 minutes ago, ollypenrice said:

Sorry, I was about to ask Wim a question then the answer dawned on me! Please ignore...

lly

Glad I could help. 

[ollypenrice]

2 minutes ago, wimvb said:

Glad I could help. 

I was going to ask how you know how much NR to give a linear image but, of course, you're looking at a preview screen stretch. However, this might not be the final stretch if this were a sophisticated stretch for pulling out tidal tails etc.

Olly

[wimvb]

37 minutes ago, ollypenrice said:

I was going to ask how you know how much NR to give a linear image but, of course, you're looking at a preview screen stretch. However, this might not be the final stretch if this were a sophisticated stretch for pulling out tidal tails etc.

Olly

Less is more, in this case. Pre-stretch noise reduction (not elimination) allows a more aggressive stretch. For my dslr images, and a number of rgb images that I processed, tgv denoise is best at preserving fine detail. Sometimes I blend the noise reduced image with the non-reduced image, 50-50.

If the objective is to reveal extremely weak signal, there's really only one way to reduce noise: more exposure.

But lately, I find I use less noise reduction in my workflow. Maybe I'm not that paranoid about noise anymore. Or could it be that a modern, cooled, mono cmos has less noise than an aging, uncooled dslr? Can't imagine that being the reason, can it?

[Adreneline]

Just to round things out I have spent the last few days taking the osc data from the ED80/414 and the lum data from the ZS71/428 and doing my best to reduce noise with TGVDenoise. (I decided to not use the Ha data as there was too little of it and it was contributing nothing of significance to the final image).

I then duplicated the DBE/noise reduced images and used one pair to fully process in PI (using the new ArcsinhStretch) and the other to fully process in PS (using the layering/blending advocated by Olly). I think I've gone as far as I can go - it is now a case of "you pays your money - you takes your pick" as to which one is better - or less bad! I like them both - but then I am biased.

This is the PI image:

This is the PS image:

Thank you to @wimvb and @ollypenrice for all your advice on processing and to everyone else who has contributed to this thread/topic.

Adrian

[wimvb]

I think you managed better with the PS version. Smaller stars and less green in the central area of the galaxy. But I'm looking at this on my mobile device, which has a small screen.

[Adreneline]

2 hours ago, wimvb said:

I think you managed better with the PS version. Smaller stars and less green in the central area of the galaxy. But I'm looking at this on my mobile device, which has a small screen.

Thank you Wim. I would really appreciate an observation on the noise. Have I overdone it again or not done enough. I've been trying TGVDenoise for the first time;I used it on both the osc and lum data.

Many thanks.

Adrian