IC443 - which image stack is best to start post processing?

[peroni]

Hey All

This week I was able to gather 8 hours of data over two evenings. My dilemma is how to select the best data for stacking and post processing.

My kit is ED80 scope guided with finder guider, camera is Atik 460ex.

2 frames were discarded due to satellite and plane trails. 

I'm using PixInsight to stack and have followed Harry's vids using the batchpreprocessing script and included the cosmeticcorrection process.

Below is the whole stack of 30 frame 15mins Ha

The first evening was clear but a little hazy and has resulted in some finer detail being lost. 

The full 30 stack results in less noise but a loss of finer detail.

Below is a stack on 15 frames from the best night.

And finally, a compromise of sorts. The best 75 percent of frames stack (23) as determined by Registax 6

So my question is...

I appreciate more is always better but, armed with 8 hours of Ha data, how do you decide which images to stack before you embark on the lengthy post processing work.

Is there a tool in PixInsight to select the best frames?

[Uranium235]

For me, the answer is no - there is no shortcut to QC on my subs. I usually load all of one filter (or pane), then partly stretch all of them so I can see which ones have a gradient/cloud/moon damage/etc... Weed out the dirt, keep the clean(ish) ones.

Actually, ive just sorted a load of Orion subs that way. Turns out I need to replace about four hours worth.

[peroni]

For me, the answer is no - there is no shortcut to QC on my subs. I usually load all of one filter (or pane), then partly stretch all of them so I can see which ones have a gradient/cloud/moon damage/etc... Weed out the dirt, keep the clean(ish) ones.

Actually, ive just sorted a load of Orion subs that way. Turns out I need to replace about four hours worth.

That's pretty much what I've done. I was hoping there was a qualitative method rather than the old 'eyeball'.

I did find the 'AutoSTF' script. That was useful to get the 30 images stretched for assessing.

[strutsinaction]

Spooky. I'm collecting Ha tonight for my first attempt at IC443. In fact it's my first attempt at Ha full stop   I'm hoping my data looks as good as yours. I've gone for 20x10m subs with my QSI690. SGPro doing its stuff as I type.

Are you planning to collect any RGB for this target?

Regards

John

[peroni]

Spooky. I'm collecting Ha tonight for my first attempt at IC443. In fact it's my first attempt at Ha full stop   I'm hoping my data looks as good as yours. I've gone for 20x10m subs with my QSI690. SGPro doing its stuff as I type.

Are you planning to collect any RGB for this target?

Regards

John

I don't have any RGB filters yet. I'm trying out the narrowband route so on the next clear night I'll grab so OIII data. It's a beautiful nebula, hopefully you'll grab some good shots.

[frugal]

I have been reading around image integration and how to determine which algorithms are best. As you are using PixInsight, have a good read of the Image Integration, especially the console statistics section. There is some useful information in there about how the S/N improves for the stack with respect to the reference image. So you could try each of the variations and record the stats for comparison.

There is also a good presentation with worked examples in one of the referneces from the documentation http://astrosurf.com/jordigallego/articles/Image_integration_JGallego.ppt which is a useful read.

It may not help you to choose whicn subs to keep, but it might give you an empirical set of numbers to compare the results of the various options.

There is also the Subframe Selector script in PixInsight which provides metrics on each your subframes so that you can chose the best of them http://pixinsight.com/doc/scripts/SubframeSelector/SubframeSelector.html

[tingting44]

just a small note, you do not need to remove subs with satellite trails in, PI gets rid of the trails, so all is good there

[peroni]

I have been reading around image integration and how to determine which algorithms are best. As you are using PixInsight, have a good read of the Image Integration, especially the console statistics section. There is some useful information in there about how the S/N improves for the stack with respect to the reference image. So you could try each of the variations and record the stats for comparison.

There is also a good presentation with worked examples in one of the referneces from the documentation http://astrosurf.com/jordigallego/articles/Image_integration_JGallego.ppt which is a useful read.

It may not help you to choose whicn subs to keep, but it might give you an empirical set of numbers to compare the results of the various options.

There is also the Subframe Selector script in PixInsight which provides metrics on each your subframes so that you can chose the best of them http://pixinsight.com/doc/scripts/SubframeSelector/SubframeSelector.html

I'm off for some more reading now. Thanks for the pointers.

[peroni]

After reading the http://astrosurf.com/jordigallego/articles/Image_integration_JGallego.ppt presentation I've become a data junkie. Trial and error today has proven to me that more frames stacked is better (yes, yes, I know everyone has been telling me  )

I've decided to discard one particularly nasty frame with a massive plane passing through. I figured this could be tricky to fully eliminate without discarding data unnecessarily. (maybe I'll try later, it's been a busy day)

I did keep the frame with a satellite in.

Moving from 30 frames to 31 frames gave me an improved SNR from 4.9138 to 5.092 (3.6% improvement). A total pixel rejection of 0.018%.

Below is the image I'll work on for post processing. I think the benefit of excluding frames to reduce the 'haze' in the image is not sufficient to sacrifice SNR. As lots of people have said before... "more data is better" 

Images were stacked in ImageIntegration using the 'linear fit clipping' rejection algorithm and sigma values of 8.5 low and 6.75 high.

[Astrosurf]

Gorgeous image. And very interesting info! 

What's that bright star? Would it be in my SW handset menu?

EDIT: found it in Astrobin: Propus!

Alexxx

[peroni]

Gorgeous image. And very interesting info! 

What's that bright star? Would it be in my SW handset menu?

EDIT: found it in Astrobin: Propus!

Alexxx

 Correct. 

Eta Gemini, aka HD42995, aka Propus, mag 3.31

not from memory, just from CdC 

[peroni]

What's that bright star? Would it be in my SW handset menu?

IC443 should be in the sw handset menu, that'll get you to the nebula.

[frostynixon]

Hello,

nice images of IC443. I`m working myself on this object at the moment. So far i`m at 5 hours integration time but especially the middle section is not as feature rich as your images.

br

[peroni]

nice images of IC443. I`m working myself on this object at the moment. So far i`m at 5 hours integration time but especially the middle section is not as feature rich as your images.

Thanks, my image is a stack of 31 x 15 mins so far. (7 hours 45 mins)

I'm getting some more frames as I type now. Hopefully another 3 hours. I'll post the new total stack once I've added them in.

[peroni]

I grabbed another 10 frames last night before clouds rolled in. SNR is up to 6.925 now (a 36% improvement on the 31 frame stack). 

There were 2 frames with plane and satellite trails in last night which I left in the stack. Subtraction of those worked ok.

Hopefully you can see the difference between the two stacks in this screenie below. 31 frame stack in on the left, 41 frame stack is on the right.

[Astrosurf]

I'm afraid I can't see a difference. Maybe this site's image compression is to blame. I have heard that there's usually not much point in taking more than 30 subs as it doesn't add much in the way of data, but I don't know!

[peroni]

I'm afraid I can't see a difference. Maybe this site's image compression is to blame. I have heard that there's usually not much point in taking more than 30 subs as it doesn't add much in the way of data, but I don't know!

Here's a dropbox link to the image. I think this is a bit clearer. My next question was going to be... how many images can I stack before I see no benefit?

We'll soon see. Hopefully there will be some more clear nights where I can grab some extra frames.

[frugal]

For any given number of subs, in order to double the S/N you need to quadruple the number of subs.

[alcol620]

Peroni

 

Where do you get the SNR number from?

[Pompey Monkey]

On 24/02/2015 at 13:37, peroni said:

Here's a dropbox link to the image. I think this is a bit clearer. My next question was going to be... how many images can I stack before I see no benefit?

We'll soon see. Hopefully there will be some more clear nights where I can grab some extra frames.

I like this thread - Thanks Peroni!

Craig Stark did a good presentation a while back. This is the second time I've posted a link to this today, but it's worth it: 

 

[Pompey Monkey]

On 21/02/2015 at 20:47, Uranium235 said:

For me, the answer is no - there is no shortcut to QC on my subs. I usually load all of one filter (or pane), then partly stretch all of them so I can see which ones have a gradient/cloud/moon damage/etc... Weed out the dirt, keep the clean(ish) ones.

Actually, ive just sorted a load of Orion subs that way. Turns out I need to replace about four hours worth.

I still have not got my rig dialled in enough to be as brutal as I need to be without the instant gratification gene taking over.... I read a thread on this exact matter over on Cloudy Nights a little while ago and the consensus was the same: No pain, no gain! I'll get there in the end