Sub selection for final images

Aries Productions Astro Pixel Processor (APP) Processing Stacking Deep Sky M31
4 replies•106 views
Bradley Watson avatar
Bradley Watson avatar
Hi all,

With weather being pretty bad right now, I have been looking at my data and have had a look at the quality of my subs. I use APP for stacking which provides some useful information, I now take the process to normalisation where I then decide which subs to keep and then integrate. The question I have for you is what criteria do you use for selecting subs, I am sure this may differ depending on stacking software?

I have been using the "quality score" (APP), but now started using FWHM, is this the BEST parameter to use? (I recently looked at my Andromeda data and have deleted 4 out of 5 hours due to new selection criteria). Do you just eyeball your images……………?

Thoughts and selection process are most welcome
CS
Brad
Engaging
Kevin Morefield avatar
Kevin Morefield avatar
All of the above really.  Usually the cut-offs are relative to the median value.  I want to eliminate outliers mostly.  But in the tough weather months I’ll have to cull much more harshly.

I initially use the eccentricity to cull.  In PI I use a cut-off of .6 (whatever that means…smile or using CCDInspector an elongation of 25%.

I hate gradients so if I visually see a gradient in a sub vs. the rest in the stack I’ll drop it.   No tool needed for that, though Blink in PI is good for this.  This and eccentricity are done in the morning right away.

FWHM and SNR culls I usually do when I feel like I’m done collecting data for the project.   To get an accurate comparison of subs I usually do this on registered subs.  In CCDStack, you can click on a star and it will give you the FWHM for that star in every sub.  I will use a much tighter cut-off for luminance than color.  The color sharpness matters a lot less so no need to be as picky with FWHM.  I pick a non-saturated star somewhere near the middle of the frame and then test a few to be sure.  CCDInspector and PI Sub-frame selector are also good tools.

SNR is harder.  And many stacking routines will weight the sub based on SNR so it may be less important to cull.  But, I worry about programs that don’t tell you ahead of time what the weighting is that will be used. In PI sub-frame selector I often see SNR weights that make no sense.  Will the image integration tool use that bad assessment too?  I like to use the normalization tool in CCDStack. It allows me to manually select an area of the background and a highlight or object area and then calculates a ratio of the difference.  So this is signal to background literally but that’s really what I want.  Those subs taken with poor transparency immediately become clear.  I usually will eliminate subs where this ratio is less than about 50% of the norm.

If all of these culls leave me with too little data in a particular channel, I get more data.  So while I said I look at FWHM and SNR at the end, I really do it all along too so I can rebalance what I’m capturing as I go.   But then I’m retired so time is cheap…
Helpful
dkamen avatar
dkamen avatar
Hi Brad,

I believe APP calculates scores and statistics at the star analysis phase which comes right after calibration. Scores do not change with normalisation.

I have noticed recently (this Monday actually smile  ) that FWHM can change a lot depending on the number of stars you set as target for the analysis phase, so especially for  wide fields a small number can be misleading as larger stars tend to have higher FWHM anyway.

I think star shape is at least as important as FWHM. If you have minor trailing or shaking, the star might be relatively sharper than a properly tracked star (hence smaller FWHM) but obviously it will not be better.

FWHM may also be misleading because dimmer images have smaller FHWM. To test, do a quick integration of your 10 best subs (based on quality) and analyse it as a light frame. You will see that its FWHM can be significantly higher, but it  obviously as overall quality the integration will beat  any individual sub hands down.

Personally, I find quality (which combines all statistics) does a very good job for comparable subs. It does get a little confused when the subs are too different (say you have 30 second ISO1000 subs next to 300 second ISO100 subs from a different night and with slightly different field of view). In that case I usually prefer to use duration. Finally, star shape is good for targets where stars are the most prominent part of the image (rich star fields or clusters).

Cheers,
Dimitris
Helpful Insightful Respectful Engaging
Bradley Watson avatar
Bradley Watson avatar
Kevin Morefield:
If all of these culls leave me with too little data in a particular channel, I get more data.  So while I said I look at FWHM and SNR at the end, I really do it all along too so I can rebalance what I’m capturing as I go.   But then I’m retired so time is cheap…


Kevin, thanks for the response.

I think you have hit the nail on the head. For me personally, and this is with only a few months of knowledge, it feels to me that I cannot entirely trust software to get what I want, its a real balancing act, like I said I have now found myself going back and looking at my data and assessing at SNR, FWHM, star shape etc. I suppose it could be good to have a good standard practice/routine for looking at this that will allow elimination of outliers quickly.
Bradley Watson avatar
Bradley Watson avatar
I believe APP calculates scores and statistics at the star analysis phase which comes right after calibration. Scores do not change with normalisation.


Hi Dimitri, thanks for responding. You are absolutely correct, APP does calculate scores during star analysis and further statistics in the steps before integration (giving dispersion, background etc.). You can actually weight this yourself but how can you really get what you want. For me, star shape now is not that important as I have a very forgiving setup (I would have to be hitting my mount with a baseball bat), but SNR (bortle 8 sky) is important along with other parameters.
With FWHM, this can vary just like you say, so you have to be careful here too.
Do you trust APP entirely or do you look at each parameter to check?
For me with a steep learning curve ahead of me, this thinking means I am focusing a lot on acquisition as this is probably one of the biggest challenges we'll face as APs and I have almost ignored this and focused on post processing.
Thanks for you input, really helpful
Respectful Supportive
Related discussions
What subframes do you keep / How many subframes?
I know this question is very subjective, but I am trying to better my processing. I will use NGC6960 (Witches Broom) as the image to discuss, as we all know it. I have collected images for the last 2 months on nights that I could. I am imaging it for...
Directly addresses subframe selection criteria, the author's main question.
Jul 22, 2022
Handy Google Sheet for FoV, Seeing, FWHM, Resolution / Image Scale, Imaging/Guiding ratio, Power capacity, etc
Hello, The past days I have constructed a Google Sheet to help me combine the right imaging camera, guiding camera, imaging scope and guiding scope, plus the necessary power required for the whole setup. This Google sheet displays some metrics such a...
Includes FWHM metrics relevant to subframe quality assessment.
Dec 2, 2021
Hello from southern Delaware, USA
Hi everyone! I have been an on/off amateur astronomer for decades; after entirely too many off years, I jumped back in a couple years ago. Last fall I decided to take up astrophotography…so down the rabbit hole I went and there has been no look...
Image processing and stacking techniques directly relevant to author's stated needs.
Sep 14, 2019