What is Photon noise?

MichailK

Mu-43 Top Veteran
Joined
Nov 6, 2017
Messages
752
Location
Thessaly, Greece
any physics majors on here?
not here but what I understand is that photons quanta on dark parts of scenes are way scarcer than on the bright parts so the noise is inherent in nature (think of geiger meter random sounds away from a radiation source vs constant white noise hum when close, as in the movies) and not affected much more by all the other noise sources of the camera’s imaging system

we do not directly see this because in contrast to a one-shot camera photo our brains’ imaging system continuously stitches multiple shots taken during constant inconspicuous saccadic eye movements, at different apertures depending on brightness of the sub-scene under capture and accumulates&interpolates dark noisy shots to be presented as noiseless to the higher brains structures that we perceive as “me seeing”

once upon a time audio engineering has taught me that, on the sound side,
the “loud” is way more energy carrying than “loud enough” which is way more energy carrying than ”ok level” which is way more energy carrying than “quiet” which is way more energy carrying than “whisper” hence we need a very not linear way to express the colossal energy difference between “whisper” and ”loud”, hence the dB scale
and I suspect in the light waves, things are not that different to the sound waves side of things along the light sensitive neurons vs the sound sensitive neurons non-linearity

but I also wonder, how many photons per surface unit per time we talk about?
 
Last edited:

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
723
There is (of course) wiki page about it.
Here is explanation more focused on photography.

Generally it's caused by the fact that flux of photons from the scene isn't constant but it's fluctuating and similarly detection of photon by the sensor isn't secured but has some probability of success (which is pretty high for current sensors).

Personally I like analogy with the rain where rain drops represents single photons.
If one has array of buckets as sensor where every bucket is a single pixel then in case of less intensive rain and short times, there will be large differences about number of rain drops collected by buckets. Those differences represents noise and in this case it will be easily visible. For heavy rain and/or long collecting times, differences between different buckets will be difficult to spot on.
Important thing is that what's make those differences (noise) clearly visible is their relative prominence and not their absolute magnitude. That's why few drops difference will be nicely visible when in average every bucket contains lets say six drops but one hundred drops difference will not be visible if average for every bucket will be ten thousands drops.
So more rain drops (photons) means less prominent differences (noise).

There is very simple equation for photon noise where photon(=shot) noise is equal to square root of signal where signal is in digital photography number of photons successfully detected by camera.
 
Last edited:

Hannety

Mu-43 Regular
Joined
Aug 15, 2019
Messages
137
It is shot noise originating from the particle nature of light.

https://en.wikipedia.org/wiki/Shot_noise

If a camera sensor is evenly illuminated so photosites on average capture 10 photons some will capture 11 or 12 and some 9 or 8. They capture a Poisson distribution and the variation or noise is proportional to the square root of the captured number. So in this 10 photon case 3.1 of the 10 photons are noise which is about 1/3rd.

Capturing 100 photons 10 or 1/10th would be noise.

Photon shot noise is the most significant source of noise in digital imaging and is unavoidable. Photon noise is predominantly why larger sensors perform better than small ones. A photosite 4 times the area captures 4 times more photons at the same illumination level and so produces half the noise. Additionally the maximum number of photons captured before saturation is proportional to the photosite area so a photosite 4 times the area can accept 4 times more illumination and so 4 times more photons halving noise again.

Full frame sensors have about 2 stops noise advantage over m4/3 and the combination of more photons at the top and less noise at the bottom gives at least 3 stops more dynamic range.
 

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
723
@Hannety FF cameras have 2 stops advantage (in case of the same technology) but I'm not sure where you took information about " at least 3 stops more dynamic range ".
 

junkyardsparkle

haunted scrap heap
Joined
Nov 17, 2016
Messages
2,440
They capture a Poisson distribution and the variation or noise is proportional to the square root of the captured number. So in this 10 photon case 3.1 of the 10 photons are noise which is about 1/3rd. Capturing 100 photons 10 or 1/10th would be noise.
To avoid giving somebody the idea that "some photons are noise and some aren't", might be better to emphasize that the "noise" is simply the random variation between adjacent collection sites, which tend to average out as predicted by the distribution model... this means not only over the duration of one continuous collection, but even if you average the pixels from two images taken at completely different times (given an equivalent signal). This is why you can stack shots for noise reduction.

Another way to think of the process would be: if you throw a handful of darts at a small marked area on a clean wall, the resulting arrangement from this small sampling might appear to be a pattern, but if you spent all day throwing those darts (and drinking enough beer to keep your aim from improving significantly) you'd end up with more or less the random distribution predicted by the model. If you had a long wall with a row of similarly skilled/impaired people doing this, over the short term there would probably be significant variation between them as to how many holes were within the target mark, but over time... not so much.
 
Last edited:

Hannety

Mu-43 Regular
Joined
Aug 15, 2019
Messages
137
@Hannety FF cameras have 2 stops advantage (in case of the same technology) but I'm not sure where you took information about " at least 3 stops more dynamic range ".
You are correct. I was confused thinking more photons captured at the top end increased dynamic range while they are actually captured at the same illumination level. Only at the bottom end is two stops less illumination required to capture the same number of photons and give the same signal to noise ratio.
 

BobBill

Mu-43 Veteran
Joined
Dec 29, 2010
Messages
317
Location
MN USA
Real Name
Bob Hively-Johnson
Might this be the same as "sensor reflection," which I recently learned about and maybe encounted.
 

Mark73

Mu-43 Regular
Joined
Sep 13, 2019
Messages
58
A good breakdown here, but you might want to gloss over the astrophotography bits, if it doesn't apply to you.

The buckets of rain analogy is also mentioned.
 

BobBill

Mu-43 Veteran
Joined
Dec 29, 2010
Messages
317
Location
MN USA
Real Name
Bob Hively-Johnson
Mark73, What is this "Buckets of Rain" thingy? Am learning and very naive re digital stuff.

I mentioned the reflection as I recently experienced a blue area on a "night moon grab." Only time I noticed and not sure what it was. Maybe a bucket of rain?

Keep it simple, which I know is not so easily done!
 

Mark73

Mu-43 Regular
Joined
Sep 13, 2019
Messages
58
Mark73, What is this "Buckets of Rain" thingy? Am learning and very naive re digital stuff.

I mentioned the reflection as I recently experienced a blue area on a "night moon grab." Only time I noticed and not sure what it was. Maybe a bucket of rain?

Keep it simple, which I know is not so easily done!
Machi has touched on it, and it is explained further about 30 minutes into the video link I posted, should you not want to view the whole presentation, in the "Shot noise" portion.
 
Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Mu-43 is a fan site and not associated with Olympus, Panasonic, or other manufacturers mentioned on this site.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Copyright © 2009-2019 Amin Forums, LLC
Top Bottom