Affiliation
None
Tue, 04/18/2023 - 09:57
If one has (e.g. due to cloud) just one good image of a target, is it acceptable to stack multiple copies of this image in order to improve SNR values of e.g. for fainter comparison stars? I tried this on a B image of R Ari and it does indeed do this.
Many thanks
Keith Robinson
No. It doesn't work. Stacking *the same* image over and over again doesn't do anything to the quality of the image in any meaningful way, and certainly it will not improve the SNR.
The reason that stacking works on multiple, different images is that by adding those images together, you increase both the signal and the noise (in absolute terms), but the signal increases faster than the noise, so the Signal-to-noise-ratio goes up. The reason is that the noise part is a stochastic process while the signal remains the same (in theory), so to speak.
Let's say you want to know if a dice is loaded. The way to find out is to roll it often and build up the statistics how many times each number is on top. If you do that 100 times you'll get a pretty good statement about the true probability of each outcome. Now if you have just the outcome of 1 roll, can you just multiply that by 100 to judge if the dice is loaded? I guess we will all agree that that makes no sense, on paper you have 100 trials but they are not independent => counting them many times is bogus statistics.
It's the same with photometry: you'd like to get an approximation of the "true" brightness of a pixel(region) despite of all the noise sources that influence your measurement. Making independent measurements and combining them will improve your certainty on the true value, but just taking the same measurement and pretending you have so many will just multiply your pixel values with the number of copies, and then when you do the average you get the original value again. Nothing gained.
CS
HB