PixInsight SubFrame Selector Advanced Section

UPDATE (02/27/2022) PixInsight has now implemented better (and very different) schemes for weighting images- specifically PSF SIgnal Weighting. Thus the problematic SNRWeight calculation no longer the only choice (and really never is the best one). So... this video is now a solving problem exercise- for an older type problem. The solution IS useful for other parameters. If you care to really understand my methodology- you will have a nice tool to control any particular unruly parameter.
UPDATE (11/13/2021) And now this video is aged more because PixInsight now has new methods for weighting (and normalization) that exist in both SFS and ImageIntegration. Yours truly pushed the developer pretty hard for these improvements! :)
I made this video in May of 2020 (more than 1 year ago). I think this is a valuable lesson- but it is all because the scale estimator in PixInsight does not give optimal results! NormalizeScaleGradient fixes all of these kinds of issues. But if you are a super-fan of SubFrameSelector- I think you will enjoy this unique approach to using the tool.
This is the last section of a series of videos on SubFrameSelector in PixInsight at AdamBlockStudios.com . It introduces an advanced concept of creatively assigning asymmetric weights to overcome common issues when using SubFrameSelector.
I discussed this on the PixInsight Forum as well:
pixinsight.com/forum/index.ph...
This video was created before the new NormalizeScaleGradient script became available. So much of the weighting can be done with ease now.... however, this video illustrates an interesting way to solve problems.
More at:
AdamBlockStudios.com
Those just starting out in PixInsight will enjoy FastTrack Training!
www.adamblockstudios.com/cate...
(see the video @ • Introducing FastTrack ... )
REVIEWS!
www.adamblockstudios.com/cate...

[@DavidSilesIT]

This is the best subframeselector tutorial I have seen. Not only for what you are explaining but because you are explaining it very well. Thank you again.

[@AdamBlock]

Just a small part of what is available on my site.

[@AstroPhotoFacts]

All your videos are the Best PixInsight explanation i have see! , many thanks for your great support Adam , From Azores Portugal :)

[@AdamBlock]

Thank you!

[@franckmalterre6087]

Thanks Adam for another great tutorial

[@AdamBlock]

This is just an added (extra) section. The main instructional videos are more explanatory and useful for typical usage.

[@stephen2615]

My brain hurts.. But seriously, I had no idea of the power of this feature. I intend to explore SFS functionality a lot more. Thanks for this video.

[@terryrobison]

Excellent video and insight into weighting images

[@AdamBlock]

Thanks Terry.

[@mrrockenrock]

Another great Adam Block video that is true to his technique of explaining why we do what we do in Pixinsight. Sign up for his fundamentals videos to learn both how and why PI is processing your images. You won't ever be disappointed!

[@rikutalvio]

Cool concept! People are using weights blindly, be it WBPP or just slapping on a formula they found online, without ever really examining the charts and graphs. I think this could be worked into something that is more foolproof than any simple weighting formula. For my data Eccentricity is a bad metric because low Eccentricity is often accompanied by high FWHM. This method gets me around that problem. SNRWeight is still problematic, though, as it really likes anything bright passing by.

[@AnakChan]

That's great work Adam. Would love to see if the PI folks would consider integrating your formula into SubFrame Selector. I noticed that you didn't use FWHM and Eccentricity as I guess this formula is targeted to weigh specifically on noise evaluation. If you're able to delve into a similar analysis w.r.t. FWHM and Eccentricity too that'll be greatly appreciated as I found (at least in my setup where I'm shooting at 2500mm FL with 3.76µm pixel pitch), I scrutinise and question the accuracy of FWHM esp vs HFR (which I get output from N.I.N.A.) and visual pixel peeping. I find that visual, esp double stars where I scrutinise the separation, tends to agree with HFR more so than FWHM, and I think Eccentricity is related to it. NB: I know FWHM and HFR slices and dices different aspects of the star and therefore as a measurement it's somewhat comparing Granny Smiths to Red Delicious, however I've just found FWHM to be less useful when used in a Weighting calculation.

[@AdamBlock]

My opinion is that resolution of images is secondary to S/N. It doesn't matter how sharp your images are if you cannot see your object (or it is too noisy to process). Eccentricity is also a secondary metric for me in general. If only a fraction of images are elongated- the combination of rejection and integration will still yield OK images. I see people outright reject these frames. If you have a ten minute exposure- and wind-shake or bad guiding make elongated stars- that tiny one second excursion means through out the frame? If the object is a faint nebula/galaxy- the small tracking errors will have no significant effect on the quality of the integrated result.

[@M31glow]

great post! This is a great lesson o how t think mathematically

[@AdamBlock]

Thank you.

[@Nico.75]

Thanks a lot for this great tutorial! One question - should I now use your weighting expression or the "PSF Signal Weight" for registering and integration? Thank you for a short feedback. :)

[@AdamBlock]

I think the answer is yes... I hope to be doing some comparisons soon.

[@christianvulpescu1398]

Just stunning, dear Adam! The only thing, I didn't get is, why not weight just by stars. Why is the punishing thing of the SNR necessary? CS Christian

[@AdamBlock]

Stars count fails under other conditions as well. There is no single metric that does it properly in my opinion. So you use more than one to confirm/modify the other and gain greater confidence in the result.

[@christianvulpescu1398]

@@AdamBlock Thank you. Now I understand. 😃👍

[@IcedReaver]

Nice video and critical thinking, even though you admitted to not being a mathematician in you PI post! Some questions. At this point, are you only looking at your best subs? Would you evolve your formula to further add weight based on the FWHM and star eccentricity values of the best subs? I ask this as I first use the approval box above the weighting box to reject subs based on their FWHM and eccentricity values in conjunction with blinking my images, and I then apply weighting afterwards, but I'm wondering if we can't use the same type scaling formula for these properties in addition to the SNRweight.

[@AdamBlock]

I am considering the full spectrum of subframe quality (from poor to good). As mentioned above, my answer today is that my primary weighting method is using star flux through NSG. There is only one weight to you assign an image with respect to ImageIntegration... so you need to pick the method to do it...

[@ExoPhotography1]

Just watched this and read your update in the video description, just to clarify - Is SFS now updated so that the SNRW shows "right" values according to how the images looks like? /Daniel

[@AdamBlock]

Pixinsight now has new measures that are improved. Whether they work to the degree necessary is still an open question.

[@ExoPhotography1]

@@AdamBlock I’ll check it next project :)

[@saritageetsangeet]

बहुत सुंदर आपके घर आ कर बहुत अच्छा लगा कल आप इसी टाइम आयेगा चाय पीने हमने 👍डन 👍कर दीया 👍

[@jualbar1]

graciassss graciassssss. perfecto. tengo una sola duda mi camara es una asi 1600 mm segun dice esta entre los 12 bits? puede ser. y asi tengo que poner? 12 bits.,? la tabla ya me dio la info el field of view calculador poniendo la camara y el sky watchwer sprit 80 da 1.45 de arco por pixel

[@AdamBlock]

Yes, in the initial "System Parameters" you should specify 12-bit if that is what your sensor is. This will affect some of the calculated values- it will be easier to understand things if you set this correctly.

[@yomichee]

Adam I want to buy one of your courses - I just started this November, spent cash on decent modest equipment and have some data to work with. I have just purchased PI and have done a couple of your tutorials posted here and appreciate your teaching method. I just don't know which one to get. I want to learn to stack properly and use the processes like the ones you have spotlighted here (baited me lol) Please advise. TIA, Michele.

[@AdamBlock]

Hi Michelle- I basically have two collections (VERY LARGE) of videos. Fundamentals is where to start- it will answer all of the questions you want to know now..and perhaps many you haven't thought of yet. :) Horizons is the subscription based additional set of videos that you can enjoy after absorbing the material in Fundamentals. I would encourage you to read the reviews and see what others say.. www.adamblockstudios.com/categories/Reviews Thanks!!

[@mrrockenrock]

Hi Adam, Is there a way to get a Fits header number to show up in Subframe Selector? I am trying to get Nweight to show up here so I can plot it vs PSF Power or vs PSF Signal. (Maybe Python, but I don't know how to use it.) Alternately if I could get the File name (which has the wXXX (Nweight) at beginning of name) in name order and index number, then at least I could see if PSF power and PSF signal are trending in the same direction as Nweight. I have a love/hate relationship with this Subframe selector tool! Like, why not have the data presented in the plot according to the sort data in Measurements table? There is a huge amount of data calculated for subframe selector, and such an inflexible plotting tool. Oh, yes, I can export a CSV file and take it into Excel, and plot to my heart's content!

[@AdamBlock]

That is what I use.(Excel) You can copy and past the output of the Console into Excel and make columns quickly...and you will note I have been very quiet on the relationship between NSG and PSF signal. I would give it a little time before doing the analysis you want to do. Likely the PSF Signal calculation will be evolving.

[@mrrockenrock]

@@AdamBlock Of course I have a big smile on my face. Going to use Nweight because I need it anyway for simplifying gradients.

[@neverfox]

In the comments of your first video on NSG, you mentioned to me "I have really never been a fan of weighting images based on FWHM or the other metrics. Too much emphasis on this! Usually the quality of an image is based on the ultimate S/N... and when you give crappy images more weight than they are worth you are hurting this ultimate result. So... I value this [NSG] signal-noise based weighting over things found in SFS." Now you're showing how to calculate weights from SFS metrics. What am I missing? Have you changed your preference?

[@AdamBlock]

Not at all. I still agree with myself! However, not everyone does and I certainly want to show different ways of thinking about things. The fact I need to bend over backwards and create some function to get what I want is just another bit of evidence in my book. :)

[@anata5127]

What about eccentricity and FWHM? I select eccentricity below 0.7 and FWHM no more than 20% of theoretical including seeing. Oh, forgot about selection by numbers of stars. I am surprised by seeing your eccentricity and FWHM. I thought your FWHM will be

[@siriusfritz4003]

The pixels must be punished!

[@AdamBlock]

Yes! When are go bad... punish 'em! lol

[@bogdanbacila9024]

frustrating way of explaning...could not follow till the end..