Back to forums

Subframe Selector Help

Pleiades Astrophoto PixInsight Acquisition Processing
3 replies•52 views
Craig Dixon avatar
Craig Dixon avatar
Hi All.

I usually just inspect each sub-exposure manually by eye before running through WBPP but I've recently started using subframe selector with the following settings:

FWHMSigma <= 2 &&

EccentricitySigma <= 2 &&

MedianSigma <= 2 &&

StarsSigma >= -1.5

This reduces the dataset considerably. I have 350 subs for my latest project and the above settings reduced this to 268. I then decided to run these settings again on the 268, which was then reduced to 206 subs. If I keep going, will it basically whittle the data set down to almost nothing? I'm worried that I'm throwing good data away.
Well Written Helpful Respectful Concise Engaging
Kristof Vandebeek avatar
Kristof Vandebeek avatar
Yes, if you apply the same Subframe Selector criteria over and over again, you will continue to remove more and more sub-exposures, potentially down to almost nothing. This happens because each iteration recalculates the statistics on an already filtered dataset, making it progressively stricter.
Well Written Helpful Insightful Concise
Craig Dixon avatar
Craig Dixon avatar
That makes sense, thanks. So my concern is if I have a large dataset and I use SFS to cull bad frames, but then go out and get more data, am I going to keep loosing data by doing so? Eg. I have a 2 panel mosaic project on the go for IC1805 and my target is to acquire 1,000 mins per filter (SHO), per panel. I've partially hit that target already but after I run everything through SFS, the number will reduce so I'll want to go out and get more data.
Well Written
Niels_L avatar
Niels_L avatar
Your settings are based on the dataset that you put in.  It's the standard deviation of the dataset. Obviously, when you use a smaller dataset, the standard deviations are also smaller but can still contain 'outliers'. I use more or less the same settings for SFS, but always look at the outcomes if it makes 'sense'. I sometimes add an upper limit for the FWHM is the dataset was not so good data. 
In principle you can calculate (through SFS) the standard deviations multiple times, but why not just make the settings tighter? E.g. FWHMSigma <1

In my opinion it depends on your dataset how you treat it.  I hope that makes sense.