Pages: [1]   Go Down

Author Topic: Dynamic Range Gains from Averaging  (Read 1031 times)

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 136
    • http://bayimages.net/
Dynamic Range Gains from Averaging
« on: February 03, 2017, 12:19:50 PM »

I've been shooting with my sony a7r2 and I've started using the Smooth Reflection app. The app averages a series of shots (from 2 to 1024 I think) to replicate the effect of an ND filter and produces a single RAW file in camera. As an ND filter, I love the application because it gives me infinite variability and removes issues of color cast, flare, vignetting. Basically it is an (almost) perfect ND.

Of course the averaging process also reduces the noise. So I'm wondering how much extra dynamic range can you get from averaging shots? Does it basically increase as log_2 (sqrt(n))? So 4 shots is 1 extra stop, 8 shots is 1.5 stops?

I suppose if noise were IID you could average it out of existence so then I'd be limited by the A/D (14bits in my body?).

From an efficiency standpoint, averaging is probably not as effective as bracketing with different exposure values. But averaging with the app has practical advantages like: (1) produces a single RAW file, (2) movement appears natural since all the exposures are the same and I don't need to mask/deghost motion, (3) I can avoid layers blending in PS or using HDR merge.

I'm actually very happy with the single shot performance of my sensor. I'm just exploring alternative ways of working and the SR app makes it easy to average with almost no workflow penalty.
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1621
    • The Last Word
Re: Dynamic Range Gains from Averaging
« Reply #1 on: February 03, 2017, 12:27:09 PM »

I've been shooting with my sony a7r2 and I've started using the Smooth Reflection app. The app averages a series of shots (from 2 to 1024 I think) to replicate the effect of an ND filter and produces a single RAW file in camera. As an ND filter, I love the application because it gives me infinite variability and removes issues of color cast, flare, vignetting. Basically it is an (almost) perfect ND.

Of course the averaging process also reduces the noise. So I'm wondering how much extra dynamic range can you get from averaging shots? Does it basically increase as log_2 (sqrt(n))? So 4 shots is 1 extra stop, 8 shots is 1.5 stops?

For noise that is frame-to-frame uncorrrelated, yes:

http://blog.kasson.com/the-last-word/an-mf-camera-in-your-jacket-pocket/


I suppose if noise were IID you could average it out of existence so then I'd be limited by the A/D (14bits in my body?).

You wouldn't be limited by the precision of your camera if you averaged at higher precision and saved the resultant higher precision image. The noise would just provide the requisite dither.

From an efficiency standpoint, averaging is probably not as effective as bracketing with different exposure values. But averaging with the app has practical advantages like: (1) produces a single RAW file, (2) movement appears natural since all the exposures are the same and I don't need to mask/deghost motion, (3) I can avoid layers blending in PS or using HDR merge.

I'm actually very happy with the single shot performance of my sensor. I'm just exploring alternative ways of working and the SR app makes it easy to average with almost no workflow penalty.

Jim

EricV

  • Full Member
  • ***
  • Offline Offline
  • Posts: 210
Re: Dynamic Range Gains from Averaging
« Reply #2 on: February 03, 2017, 02:10:26 PM »

From an efficiency standpoint, averaging is probably not as effective as bracketing with different exposure values.
 
Your equation is correct.  So for dynamic range and noise improvement, a single shot at +3 stops is worth 64 averaged frames. 

Once bracketing or averaging get you past the bit limit of your image file, you need special techniques like HDR to preserve the dynamic range and noise improvement.
Logged

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 136
    • http://bayimages.net/
Re: Dynamic Range Gains from Averaging
« Reply #3 on: February 04, 2017, 12:24:37 PM »

Thanks for the feedback

You wouldn't be limited by the precision of your camera if you averaged at higher precision and saved the resultant higher precision image.

Right, that makes sense. I suppose the smooth reflection app must be keeping a running count in higher precision.

 
Your equation is correct.  So for dynamic range and noise improvement, a single shot at +3 stops is worth 64 averaged frames. 

I normally wouldn't do averaging as a regular practice because of shutter wear but if sony ever gets silent shutter + uncompressed raw as an option in the app I might use it more often.
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1621
    • The Last Word
Re: Dynamic Range Gains from Averaging
« Reply #4 on: February 04, 2017, 01:17:12 PM »

Right, that makes sense. I suppose the smooth reflection app must be keeping a running count in higher precision.

I normally wouldn't do averaging as a regular practice because of shutter wear but if sony ever gets silent shutter + uncompressed raw as an option in the app I might use it more often.

Sony a7x silent shutter operates with 12-bit precision always, even with uncompressed raw files. So that's not something the app is going to be able to fix. However, on the a7RII, with ISO more than about 320, there's enough read noise to properly dither a 12-bit ADC.

Jim
Pages: [1]   Go Up