Additive Stacking?

[dpaul]

One more question -

I have a large scope (30'') and the Atik Horizon CMOS mono camera - this is great camera for short exposures with low noise.
My scope is unguided on an equatorial platform so I can usually get up to 20 seconds decent tracking. With my large aperture this perfectly adequate for LRGB frames.
I've also tried with narrow band and results aren't too bad, managing to push the limit to about 25 seconds, just, but more would be nicer.

So here's my question -
I appreciate additive rather than average stacking when integrating frames is not a great idea from a S/N perspective. However if such a method is possible when image processing then I could get better results with these short frame lengths. With video astronomy additive stacking is quite common (great for outreach) but can Pixinsight so this in post processing?  If not I still think it could be a useful tool - hoping its possible!

Of course another option is 4x4 instead of 2x2 binning but I prefer the 2x2 pixel resolution.

Thanks

David

[pfile]

what would be the difference between additive and average stacking?

rob

[dpaul]

Hi Rob,

What I mean is effectively turning a single 20 second frame into a 60 second frame if 5 additive stacks of 20 second are done. This definitely works with live video astronomy where you can see the image build infront of your eyes as live stacking occurs.

David

[pfile]

i think you are asking if PI can monitor a folder and automatically stack files that are incoming from some camera... is that right? if so i don't think it can do that.

i guess why i am confused is that averages are just sums divided by the number of frames, so adding and averaging are actually the same operation (the division just scales the result, effectively just moving the decimal point in the answer.)

rob

[dpaul]

Hi Rob,

Yes I meant can it add incoming files - but not live,  just for post processing. So if I have a load of frames in a folder, it would be great if these could be fed into pixinsight and added (no averaging).

David

[RickS]

There's no difference between adding and averaging if you're going to stretch the data anyway (apart from the possibility of overflow when adding.)  Either way gives you the same SNR with just a linear scaling difference.

Cheers,
Rick.

[chris.bailey]

David

PI cant do live stacking but being the impatient sort of person I am and as there is not much else to do whilst the images come in I tend to set my imaging system up so images are mirrored to my NAS (also acts as an immediate backup). I sit on my imaging computer with PI open and have ImageCalibration, StarAlgnment and ImageIntegration processes open on the screen with Overwirte Existing Files ticked. Its just a matter of of loading up the images in the relevant NAS folder into each process and hitting run. Takes 20 or 30 seconds to get a first pass image of whats going on.

Chris

[dpaul]

Chris,
Thanks for that info - I actually take a walk around the garden for some exercise, works well!


Rick
Thanks for the note, that's a real eye opener. I reviewed two saved images, one was 30 seconds, the other 3x30seconds additive. You are right after stretching the detail looks similar but without stretching the 30 second ''appears'' to possess much less detail.

David

[ChoJin]

the 30sec has a worse SNR than the 3x30sec. What Rick is comparing is 3x30sec additive stacking vs 3x30sec averaging stacking, which is the same by just a factor (1/3 in this case): (I1+I2+I3) vs (1/3)*(I1+I2+I3).

In other words, there is no benefit for SNR to do "additive stacking" instead of "averaging stacking".

[dpaul]

Hi Cho

I agree!

David

[bonz0]

What's the point of additive stacking? Sure, your noise will still tend towards zero (assuming it's Gaussian centered on 0), but your signal will max out - your stars will get blown out.

[Carlos Milovic]

Actually, this is not true. When you add two images, noise is also increased. Always. When two Gaussian distributed signals are added, the variances are added also. Since the signal is "doubled", the SNR becomes: 2*signal / sqrt( 2*var), and here comes the famous sqrt(2) improvement. Averaging and adding together is the same thing because both the signal and noise (standard deviation) are scaled by the same factor.