Back to forums

Stack more or stack best?

Processing Stacking Deep Sky M106 Messier Objects
12 replies•607 views
Omiros Angelidis avatar
Omiros Angelidis avatar
Hello fellow APs!

I have a simple question. What will end up in a better result overall? Stack all the light frames even if some are slightly out of focus or stack only the best ones? 

I am collecting data on M106 and some of my frames were taken at low altitude 30-35 and some were slightly out of focus.

Would it be better to discard those all together or even if they are not the best they still contain data which will be of some benefit to the stacked image? 

Summing up, what will be better? Adding an hour of medium quality frames or throwing those exposure to the garbage bin?

thanks as always for your great support and knowledge!!
Respectful Engaging
Omiros Angelidis avatar
Omiros Angelidis avatar
andrea tasselli avatar
andrea tasselli avatar
SNR will be improved, resolution not so much…
Oscar avatar
Oscar avatar
Omiros Angelidis:

Which one is the defocused one? I think the bottom one has slightly bigger stars, so I'm guessing that's the one? If it is, then I think the difference in sharpness between the focused and defocused are too miniscule to affect the sharpness of the final image.
Geoff avatar
Geoff avatar
Try both. Measure both end results—FWHM, eccentricity or whatever. See which is best.
Geoff
Scott Badger avatar
Scott Badger avatar
Unless you’re out of focus to the point of donut stars, I don’t see much difference between poor focus and poor seeing. Focus should always be as good as possible, of course, but when culling subs according to some fwhm limit, I don’t think it matters whether it’s seeing or focus that drives that number.

Also, the couple times I’ve compared culled and all-in stacks, all-in was as good or even better. I do Blink first, and remove anything that clearly has nothing to offer, but with the right weighting and rejection algorithms, I think you can leave it mostly up to the software to decide the value of each sub, and on a pixel by pixel basis.

Cheers,
Scott
Well Written Helpful Insightful Respectful Engaging Supportive
Monty Chandler avatar
Monty Chandler avatar
Run your calibrated subs through subframe selector to see what the anomalies actually are.  Those two don't appear bad..
Brian Schumacher avatar
Brian Schumacher avatar
As a general rule I tend to work towards the better-quality images.  I use a three-step method, 1) visual 2) dss point system 3) then follow up with a grading technique that I use in PixinSite. To be clear I do my stacking in Pixinsite. To me it really boils down to how much data I have to work with. And to be honest I really haven't done any bench testing for comparatives to tell you one way is better than the other. It's all really subjective and become personal choice
HR_Maurer avatar
HR_Maurer avatar
My two cents: It is not possible to give a general answer. Depending on how many subs you stack, you would obtain more improvement by adding more subs, or by removing out-of-focus ones. As a general rule of thumb, i try to stack at least 20  long-exposure subs per channel (mono), if i can i stack 32 or more. With only 20 there will sometines remain some artifacts if there are satellite trails in the image.
So, if there are such satellite artifacts still present in the image, you would probably also introduce artifacts from out-of-focus stars by adding a single out-of focus frame to your ensemble. And since usually all stars in a sub are out of focus, adding multiple such images will quickly have negative effects. However, depending on your motiv, you could try stacking only the best image for the stars, stacking all images with stars removed, and then blend the stars back in.

CS Horst
Helpful
Monty Chandler avatar
Monty Chandler avatar
I set targets for each project.   M106 would be a 10hour project for me.  Regardless of how many nights I would gather 10 hours of usable subs.  Each morning I run all of my lights though PixInsights  Image Calibration; Cosmetic Correction; and Debayer processes, followed by a blink view of every sub.  I discard those I don't like.  Airplanes, wonky stars and clouds get moved to the junk folder.   I'm pretty aware of focus as I'm an active participant in my photography.  Running two rigs in a forest will keep you busy.  At least 5 - 6 total targets per night.
Helpful Engaging
Omiros Angelidis avatar
Omiros Angelidis avatar
Thanks guys for all your input!! I need to increase my total time on each target! I stacked in the M106 about 110 frames of 180 sec. The result while not outstanding its getting better. I will try to focus on the target for another couple of nights. Also I need to start my exposures from 40o and upwards.
dkamen avatar
dkamen avatar
Consider an image as consisting of various levels of detail, much like wavelet transforms do. There will be features 1 pixels wide, 2 pixels wide, 4 pixels wide and so on.

When a sub is blurred, the smaller layers are non existent but the bigger layers still are present. The sub will contribute to making the integration better for features e.g. 4 pixels wide and above while zero (or worse, wrong) information will be added on levels 1 pixel and 2 pixels.

Consequently, if you scale the image down sufficiently for the differences in sub blurriness to become invisible, the integration on that scale only improves with more subs.

For example if you scale the image at 25% zoom, it could well be that a blurry sub (or a sub with tracking inaccuracies) is visually indistinguishable from a sharp, perfect one. On that scale, the integration will almost certainly improve by including the bad subs.

The integration at 1:1 will probably worsen a bit, although outlier rejection algorithms can get you a long way assuming enough good subs.

cheers,
Dimitris
Well Written Helpful Insightful Engaging
Die Launische Diva avatar
Die Launische Diva avatar
And since our greatest enemy, light pollution, resides in the bigger wavelet layers, by having more frames will let us model and subtract light pollution gradients more accurately smile