Pixinsight Process Console Statistics

[Adam1234]

Hi all, 

I found this interesting PowerPoint presentation made by Jordi Gallego in 2010 - 'Image Integration Techniques: Increasing SNR and outlier rejection using Pixinsight' (http://www.astrosurf.com/jordigallego/articles.html)

In the presentation he describes how to use the 'Average SNR Increments' statistic in the Process Console to fine tune the rejection process to maximise SNR while keeping the desired outlier rejection level when integrating. 

 

However it seems that this term is no longer being used, what statistic should I be looking at instead, either from the process console or elsewhere within PI? I don't think SNR estimates is the right one as this seems way too high, and said number increased when I applied a rejection algorithm (ESD) compared to no rejection and based from what I have read I would expect the SNR to be highest when there is no rejection.

It seems like this term has not been used since at least 2013 as I found a post on the PI forum from 2013 where someone else asked the same question but received no responses. 

These are the stats currently displayed in the Process Console after Integration:

 

Thanks

Adam

Edited December 30, 2020 by Adam1234

[wimvb]

If you use no pixel rejection, all stack image pixels are used to calculate the average signal, which will be your integrated image. As long as the individual pixels behave nicely ( eg gaussian distribution), the integrated image will have the lowest noise possible. This is because if you reject pixels, the stack is made up of a lower number of samples (for each pixel), and the average of fewer samples will show a larger variation (more noise). But, if there are outliers, exclusion of those pixels will result in less variation = lower noise. The noise estimates, such as SNR, should go up. Ideally, those outliers would be caught during calibration (hot and cold pixels). But nothing is ideal, and pixel rejection is needed to refine the calibration, as well as remove satellite trails, cosmic ray bursts, etc.

I think you can safely use the SNR estimate, as you will be comparing numbers for different rejection settings. But also look at the rejection maps (especially high rejection). If you start to see details from your main target in those, you have a too aggressive rejection.