DSS - how do you choose a cutoff for good vs bad frames?

[rnobleeddy]

The last week ha been a mixed bag of weather. I left the kit imaging but cloud periodically ruined subs, sometimes just 1, sometimes 10 in a row.

I have 180 frames in total to stack in DSS. About 10 are totally clouded out. The rest get scores ranging from 0 to 5900, with a large number around 4000. 

I know it's a trade-off between SNR and quality - but does anyone have an approach to choosing the cutoff scientifically? 

I tend to work something out based on the data - I remove any low outliers, or if the scores are more normally distributed, I'll pick perhaps the best 80%. I will err on the side of more data. 

 

Obviously I could (and might) try stacking different numbers and seeing what happens. But I also wonder if the computer couldn't do this for me - rather than stacking the best n%, software could work out the # of images that resulted in the best stacked image. You'd just need a good measure of quality and a lot of CPUs!

 

 

 

[barbulo]

Good question. For instance, Autostakkert has a graphical representation so you can decide the percentage of subs to stack. 
Could be an interesting feature to be implemented in following versions of DSS. 

[ONIKKINEN]

How did you capture the data?

DSS score can be a bit arbitrary IMO and not a good representation of quality. I had an issue where clouded subs had unusually high scores - probably because DSS thought that was signal. I think the number of stars is more telling of the quality. Often high scores and high number of stars goes hand in hand but not always.

Number of stars, guiding RMS error and HFR are what i use to vet my frames. I dont inspect the individual frames at all and just rely on these stats to do the job for me.

I use NINA to capture and have set the filename to include the afore mentioned 3 statistics. Very easy and quick to just go through the list of subs at a glance and delete the ones that are not good enough, i go through 500 subs in a few minutes this way. Of course youll have to decide for yourself what is good enough, but i usually try to have my number of stars for all the frames be within 30% of each other. Of course if the conditions are bad i might be more relaxed with the criteria.

[tomato]

This thread might be of interest: