Quality Score unreliable for useable subs

@applal You forgot the screen captures?

I was waiting for permission to upload, as stated in the user upload data guidelines

Hi Alan @applal,

Please share some screenshots, for upload of images on the forum, no permission is needed. We give permission to upload whole datasets to our server for testing 😉

Mabula

Here are two screenshots of adjacent after sorting subframes

 The quality score is nearly the same but the quality is significantly different. 

Hi Alan @applal,

Thank you very much for the screenshots, indeed, those are big differences !

In terms of the stars in both images that are visible, it could lead to this happening. This problem should be detected after 5) Normalize in differences in sky background/dispersion and noise/Signal to Noise Ratio. Did you check how the scores relates of these 2 images after 5) Normalize?

Mabula

The error "may" be in the way I am processing. I am using a ZWO 2600 MC-pro. Typically with a Optolong Ultimate filter when shooting most nebulae. This combination is usually so clean that I find no need to use calibration frames (unless for flats to calibrate out any dust) So my work flow is typically just load all my lights and select  the Integrate (* and I will just move the slider to only integrate around 75%) and let the program run, which, I assume proceeds step by step through all processes 1-6. If I don't like the result at that point I will go back and begin to check each subframe that was used, and I will usually find one or several that need to be left out of the process, delete them, and go again. What I have noticed is that the subframes I have to get rid of are not necessarily the lowest quality score, and in some instances are very near the top of the quality score. 

IS the problem of my own doing by not having calibration frames? 

Dear Alan @applal,

Apologies for the late reply.

Okay, well that could explain it partially, if you do not use calirbation frames, than you will always have some vignetting in your data which means the sky background will not be nice even without any light pollution. This means that calculations like SNR, sky background, and dispersion will be off... They assume that the sky background is corrected for vignetting of your optics.

Mabula

My one thought through all this is to wonder if there is and we are using Stars as a comparative qualitative ranking, could not the quality rating be toggled to judge 10 to 50 stars for pinpoint characteristics. I do not doubt that this might take longer, but it would be, IMO, quite useful.

Hi Alan @applal

Thank you very much for your suggestion.

At the moment, APP collects analytical data at each step, 3) to 5), your issue should be detected with the analytical results collected at step 5) I would think 😉

Mabula

Thank you for your attention to this and your reply. 

Reading your reply stating that the analytic results collected at step 5... caused me to stop and wonder, why then is Analyze Stars at step 3? And correct me if I am wrong, but when I choose to register and select the slider to something less than ALL of the subframes, is it not just eliminating based on the Analyze Star Ranking? And if that is the case, moving the order to have Analyze Stars from 3 to 5 as the final step before Integrate seems to me to improve the process. 

Dear Alan @applal,

That is simply not possible, analyse stars always needs to be done before registration, otherwise registration is not possible, we need to know the star centroids to be able to align/register images. And once the images are registered, normalization should take place, not before registration 😉

If you need to remove bad frames because they cleary affect the integrated result, the advised workflow will be to plot the analytical data after each step 3), 4) and 5) to check which frames are clear outliers after those stages. Step 5) is the most important step, (you want to make graphs of noise, sky background, quality to really see what is going on in your dataset), because everything is combined in step 5) and those combined results are the basis for which frames will be integrated if you use the % slider in 6) Integrate.

Please let me know if this helps 😉

@applal

Do you use the graphical analysis that comes with the latest version? It allow a very useful way of clicking on a sub and removing it from the integration, and while the quality scope may not be different in the two examples you showed, other parameter such as location, background, star density etc. might be and you should be able to weed out unwanted subs. I woudl recommend keeping full calibration for each sub in any situation, keeps your field limited to sky conditions per sub at least.