I did some tests with the stacking settings in Siril, and then checking the noise values afterwards. Trying both average and median, along with the many rejection schemes and their settings, I have found the following settings to work best for me:
The sigma rejection scheme seems to give me the best of both worlds: it includes enough values to reduce noise, but seems to effectively remove anomalies such as satellites or bad guiding. The parameters are standard deviations, and adjusting the high value may be needed to remove anomalies while keeping as many significant values as possible. 2.5 is the value I tend to start with, and I do a few tests while adjusting the values until I get the lowest noise value without any artifacts.
As an example, here is the noise for a single subframe:
11:50:12: Background noise value (channel: #0): 229.406 (3.501e-03) 11:50:12: Background noise value (channel: #1): 231.793 (3.537e-03) 11:50:12: Background noise value (channel: #2): 407.668 (6.221e-03)
And here is the noise from 14 stacked images using the above settings:
11:53:20: Background noise value (channel: #0): 11.797 (1.800e-04) 11:53:20: Background noise value (channel: #1): 11.604 (1.771e-04) 11:53:20: Background noise value (channel: #2): 22.598 (3.448e-04)
