Subframe Selector: Registration errors and ROI, weighting expressions

[astrovienna]

I had read that registration errors could throw off PI's quality estimates, but I hadn't really understood the scope of it until today.  The FWHM and SNRWeight figures in an 18 subframe set were wildly different between the calibrated and registered frames, even though the actual registration errors were pretty small. The frame I used for registration went from being one of the best in the stack to being one of the worst, in both FWHM and SNRWeight. So it looks like weighting before registration is going to become my new standard.

I don't think ROI is working in the new SS process.  I figured it would let me exclude the registration edges, so the quality estimates on the registered frames would pretty much match the unregistered frames, but I got the same quality estimates with or without ROI.  Any idea what might be wrong?

One extra question:  I used (SNRWeight/SNRWeightMedian)/(FWHM/FWHMMedian) as my weighting expression.  Any thoughts on that?  I'd like to use a standard weighting expression rather than having to do a unique one for each data set.

Kevin

[Juan Conejero]

Hi Kevin,

I assume you refer to black or very dark image borders caused by lack of overlapped data after registration. In such case, our robust noise estimators should not be affected by these regions, which should be rejected as outliers. If this is not the behavior you are observing, can you please upload a data set where we can reproduce these issues?

[astrovienna]

Thanks for taking a look at this, Juan.  I uploaded 18 calibrated Ha frames to Dropbox:

https://www.dropbox.com/s/mxebacxapu7c4k5/Sh2-106%20Calibrated%20and%20Cosmetized.zip?dl=0

Kevin

[astrovienna]

Juan, did you get a chance to test that data set?  I've been checking this as I've been processing new sets of data, and I'm consistently seeing that SNR and FWHM estimates vary dramatically between pre-registration  and post-registration images. 

Does this issue also affect the noise evaluation weightings done in the ImageIntegration process?  If so, does that mean that, in practice, SubframeSelector always needs to be used, or weightings will be inaccurate if there are registration edges?

Hopefully I'm just doing something wrong in SubSel.  Thanks for your help tracking this down.

Kevin