Pixinsight Star Alignment etc

[groberts]

As a beginner with PI I'm working my way through the excellent Light Vortex tutorials + with reference to Warren Keller's book and have reached the Star Alignment section i.e. all subs have so far been calibrated, corrected etc. processed only with reference to others of the same wavelength L-R-G-B.   

In order to enforce what I'm reading I've been processing my own LRGB data I already have, which up until now has been wavelength by wavelength (see above).  Now I'm at the alignment etc. stages but it's not clear to me where I start to process the subs together to ensure they are all aligned + registerted for subsequent integration; hitherto my experience is with Deep Sky Stacker so my preconceptions of how PI might deal with these stages is somewhat biased.  So, assuming there are parallels at all - at what point herein do I start to process all the subs together, except of course for stacking itself?  

Thanks 

  

[Wyvernp]

The way I do it is to use one of my subs (say the first red sub) as the alignment sub for each of the L, R, G and B integrations. That means that all of the subs will align with that one.

[groberts]

Thanks but I'm still confused.

I currently have LRGB subs calibrated, corrected and weighted in PI.

In the next step, Star Alignment, do I either:

(A) Put all the aforesaid LRGB subs into Star Alignment + choose one of the best subs as a reference and run? or

(B) Put only each separate set of wavelength subs e.g. Red and then run + then move onto to the next wavelength and do the same again?    

Logically + experience with DSS suggests that all the subs are processed in Star Alignment with reference to a 'best sub', whichever wavelength that might be i.e. Case A? 

   

 

[Hughsie]

I have all my light frames for each lrgb channel saved in separate folders. See the website Light Vortex for an example of a good file system.

 Then use your Option A. Put them all in and align them to your best weighted lum image.

Dont select an output folder, PixInsight we automatically return the aligned files back to the folders you stored the weighted images in.

Hope that helps.

[groberts]

Thank Hughsie, it certainly makes sense. 

Just to be clear - if I do this then obviously all the subs (LRGB) end up back where they were filed but now aligned?  Presumably, if I wanted to preserve the subs as they were for some reason, I could set a new output directory and the said now aligned subs will instead go there for subsequent stacking?

Graham

[Wyvernp]

16 minutes ago, groberts said:

Thanks but I'm still confused.

I currently have LRGB subs calibrated, corrected and weighted in PI.

In the next step, Star Alignment, do I either:

(A) Put all the aforesaid LRGB subs into Star Alignment + choose one of the best subs as a reference and run? or

(B) Put only each separate set of wavelength subs e.g. Red and then run + then move onto to the next wavelength and do the same again?    

Logically + experience with DSS suggests that all the subs are processed in Star Alignment with reference to a 'best sub', whichever wavelength that might be i.e. Case A? 

   

 

A. is the way I do it

Choose best sub and align all the wavelengths to that sub. Then you can just combine the wavelengths with no need to do any alignment again.

[groberts]

Thanks, I think I'm having a 'senior moment' but it was not clear in either the otherwise excellent Light Vortex or Warren Keller + from DSS that's the route I would expect.

[groberts]

Great + thanks again - so far so good! 

The Registration worked fine but (as advised by Light Vortex) the next Local Normalisation step is unclear.

Are the files used in this process, including the reference file, the original calibrated subs or those just generated by the Star Alignment process? 

Edited June 1, 2020 by groberts

[x6gas]

Really it's what ever you find most convenient from a file management point of view and how fed up you get of repeatedly selecting common files (like calibration masters or the file you are aligning to).  You can do the same calibration, cosmetic correction, etc on all of your subs if you want but obviously when you get to the stacking process you just stack the ones for each filter separately (unless you are creating a false luminance but that's another matter...)  Maybe try different ways and see what suits you?

Personally I've settled on the following:

All subs together: Calibrate -> Cosmetic Correction

Then when I do subframe selection I load up the subs from each filter and when I output files I save each filter to a separate directory.  This is so that I can compare like with like and get an acceptable number of subs for each channel (I want similar numbers in each stack so that the noise in each is similar, though of course this depends on the subs themselves).  In case it's of use my starting criteria is SNRWeight > 0.72 and Eccentricity < 0.6 but that's partly because the short nights mean I have a lot of subs with poor SNR at the moment...

Then I register (star align) the files in each folder to my best R sub.  I also save these in their own folder.

Then I stack the files in each folder - and here it is definitely quicker to have the files in a separate folder if you are loading drizzle and / or local normalisation files.

Now as far as local normalisation goes, I believe you should normalise to the best image from each filter - so you normalise your red subs to the best R image, the green subs to your best G image etc.  AFAIK you should NOT normalise all of the subs to, for example, the sub you used to align on.  This again is an argument for having your files organised in folders by filter at that point.

However, I've abandoned local normalisation.  Warren Keller advises to use it with care - you need a very clean reference image, I believe, and I found it does no better than using "additive with scaling" which is one less process.

HTH, Ian

[groberts]

Thanks Ian for your extensive and very helpful thoughts. 

I'm edging slowly towards getting my first stacks in PI but it's been something of a journey, which I hope is (a) worth it and (b) gets better / quicker with practice.  It's these sort of workflow issues that can imo make a big difference.

Graham

[x6gas]

It takes me a LOT longer to do the preprocessing in PixInsight than it does in AstroArt (which was my 'goto' until recently).  I've yet to do a comparison of the final result to see if it is worth it.  Of course there is a script that automates much of it but I am still at the stage where I am trying to really understand PI rather than following recipes.

Cheers, Ian

[groberts]

2 hours ago, x6gas said:

It takes me a LOT longer to do the preprocessing in PixInsight than it does in AstroArt (which was my 'goto' until recently).  I've yet to do a comparison of the final result to see if it is worth it.  Of course there is a script that automates much of it but I am still at the stage where I am trying to really understand PI rather than following recipes.

Cheers, Ian

I'd be interested to see how they do compare + any other observations or comparisons with other pre-processing software regarding ease-of-use and results. 

An initial view of the stacks from PI look promising but is it worth this enormous effort - we shall see?

Graham  

Edited June 2, 2020 by groberts

[x6gas]

Well I don't know how valid these comparisons are but I stacked the same frames in PI (using sigma clipping) and AstroArt 5 (using Sigma average with the default sigma of 3; AstroArt takes a little over a minute to calibrate, register and stack the subs).  Measured using sub frame selector:

Here are tifs of the two resulting stacks if you want to take a peek.

R 45 x 300s AA.tif

R 45 x 300s PI.tif

[groberts]

Hmm interesting and if the numbers are valid they indicate that AA does a good job and is much easier + quick to use.  

I've now tried this on my own results below:

1. Original DSS stack
2. PI-1 Stack is without drizzle
3. PI_2d Stack with drizzle

Which seems to me to be unclear except for the SNR with drizzle?

I know this is my first PI pre-processing and I'll get quicker + a better feel for the settings but I have found it to be very fiddly and very time consuming.  The $64,000 question is - what is the easiest + best result software for pre-processing, clearly AA seems to be a contender? 

Graham