Author Topic: [ALREADY DONE] Read Noise E-  (Read 1404 times)

calypsob

  • Freshman
  • **
  • Posts: 57
[ALREADY DONE] Read Noise E-
« on: September 27, 2017, 05:50:58 PM »
 I would like to figure out if magic lantern can give me an evaluation of noise in my images.  Particularly with astrophotography, if I can figure out the Read e- noise value in an individual image, I can make a rough estimate of how many subs I will need to shoot for integration and also how long of an exposure I need to take in order to swamp the read noise.  This of course would need to be a reading from the raw file and not the jpeg preview and it would need to give a value listed in the R,G,B channels independently.  Cutting out the need for a computer for this process would be huge.
Full spectrum T2i
T3i

a1ex

  • Administrator
  • Hero Member
  • *****
  • Posts: 12420
Re: Read Noise E-
« Reply #1 on: September 27, 2017, 06:04:37 PM »
It can - try the raw_diag module.

There are a couple of methods:
- from OB areas (not reliable, just an extremely rough approximation, but works on any image)
- from one dark frame (includes both fixed and random components)
- from the difference of two dark frames (you'll get the random noise component * sqrt(2), assuming it's Gaussian)
- from the difference of two regular images (so you can estimate the noise at various signal levels - enough information for plotting a SNR curve).

I have some trust in the last method, especially for DR measurements, as long as white level (clipping point) is autodetected well (sometimes it isn't). However, fitting FWC and read noise from the SNR curve is probably an ill-conditioned problem - if you can suggest a better method, I'm all ears.

calypsob

  • Freshman
  • **
  • Posts: 57
Re: Read Noise E-
« Reply #2 on: September 27, 2017, 09:59:53 PM »
It can - try the raw_diag module.

There are a couple of methods:
- from OB areas (not reliable, just an extremely rough approximation, but works on any image)
- from one dark frame (includes both fixed and random components)
- from the difference of two dark frames (you'll get the random noise component * sqrt(2), assuming it's Gaussian)
- from the difference of two regular images (so you can estimate the noise at various signal levels - enough information for plotting a SNR curve).

I have some trust in the last method, especially for DR measurements, as long as white level (clipping point) is autodetected well (sometimes it isn't). However, fitting FWC and read noise from the SNR curve is probably an ill-conditioned problem - if you can suggest a better method, I'm all ears.

Alex, this sounds awesome.  I will need to check this out later on and see what I can explore the feature more
Full spectrum T2i
T3i