4
\$\begingroup\$

It's obvious that higher ISO affects the photograph by adding noise or grains.

But will that grain or noise be distributed equally over all parts of the photograph, or will it specifically affect specific colors or so?

Follow up question: Why is the blue channel the noisiest?

\$\endgroup\$
4
  • 1
    \$\begingroup\$ It brings a tear to my eye all the answers are about amplification and such things sniffle en.wikipedia.org/wiki/Film_grain \$\endgroup\$
    – Dreamager
    Jan 24, 2012 at 14:04
  • 1
    \$\begingroup\$ The answers so far seem to focus on What is “ISO” on a digital camera? / How is ISO implemented in digital cameras?. But I think this is more along the lines of Why is the blue channel the noisiest? (with that question basically being part of the answer, although I'm not clear if the question also means to ask about color fidelity or dynamic range). \$\endgroup\$
    – mattdm
    Jan 25, 2012 at 22:47
  • \$\begingroup\$ Which of course raises the question: vivek_jonam, what do you mean by "affect specific colors or so"? \$\endgroup\$
    – mattdm
    Jan 25, 2012 at 22:48
  • \$\begingroup\$ @mattdm i meant of asking whether the amount of grains in the same photograph at high ISO differs for various color?? \$\endgroup\$ Jan 26, 2012 at 1:59

4 Answers 4

8
\$\begingroup\$

When you change the ISO value to a higher, you really change the amplification in the chip.

Let's look at one single pixel first.

During exposure the pixel receives a number of photons, which generate (let's say) 100 mV, and the chip's noise gives 10 mV. You have a signal-to-noise ratio of 10:1.

Now, you need to expose half the time, and therefore you change to an ISO of the double value. Then you only get photons enough to generate 50 mV, but the noise is still 10 mV. To get the same level of signal to the in-camera processor, your chip has to amplify the signal from 50 to 100mV — but then the noise is also amplified, from 10 to 20mV. This gives a signal to noise ratio of 5:1.

This means you have the double amount of noise in your RAW data.

Normally noise is seen mostly in the darker areas of a picture, and you have to remove it, either with some software, or manually.

\$\endgroup\$
3
  • \$\begingroup\$ The noise you see when amplifying a weak signal by upping the ISO isn't primarily from the chip, it's photon noise from the random nature of light, the less light you receive the more random it is. The whole point of amplifying on the chip instead of in software is you amplify the signal produced by the photos before the read noise is added. \$\endgroup\$
    – Matt Grum
    Jan 26, 2012 at 11:04
  • \$\begingroup\$ This seems self-evident, but do you mean to imply that shooting at ISO 3400 is 34x more noisy than ISO 100? \$\endgroup\$
    – Drew
    Aug 29, 2012 at 7:03
  • \$\begingroup\$ Andrew, I'm going to go out on a limb here, but I think the ISO ratings do indeed work that way: since each doubling of ISO doubles the sensitivity of the "film", an ISO of 3200 has been doubled 5x from 100, so the noise has also doubled 5x over, from (let's say) 1000:1 to 1000:32. Noise, blurriness, or darkness: pick 1! \$\endgroup\$
    – SG1
    Aug 29, 2012 at 13:34
4
\$\begingroup\$

ISO by itself affects a photo by making it brighter. That is all.

It does not add any noise or grain (except for an inperceptible amount introduced by the amplifier). The ISO setting amplifies the analogue signal before it is digitised. This actually reduces noise you'd see compared to amplifying after digitisation (as this would amplify the digitisation noise also).

Raising the ISO setting with the camera set to a program exposure mode will cause the shutter time and/or aperture to be reduced, causing a darker photo (which is made up for by the fact that ISO makes the photo brighter). This is where the association with ISO and noise comes from.

Reducing the amount of light coming through the lens increases noise. Photons are emitted randomly by a lightsource, if you capture many photons the randomness evens out, collect fewer photons and the randomness shows up.

With this in mind, the noise introduced by lack of light affects each colour channel differently. The blue channel is the noisiest as the blue filters in the bayer matrix take out more light than the red and green. Also there are twice as many green pixels as red and blue which evens out the noise. Finally blue is also less common in nature (hence the reduced number of blue filters) so in a particular scene you will capture less blue light so the blue channel will appear noisiest, especially if you shift the colours in an image, which is akin to digitally amplifying the noise in the blue channel.

The noise will also show up more in shadow areas as these naturally reflect less incoming light. If you are working in really low light (and compensating by raising the ISO a lot) then you can see significant noise in the highlights also.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ Regarding how ISO setting amplifies or reduces noise...I think thats very much a brand-dependent thing. Canon, for one, achieves ISO both pre- and post-amplification, depending on whether you are using a lower base setting, a "pull" setting, a "push" setting, or a very high ISO setting. Its not quite as simple as the ISO setting always amplifying before digitization...it can be more complex than that. I'll see if I can dig up the reference that explains Canon's approach in detail, but suffice it to say, ISO settings can indeed have visible and detrimental effects on noise even at low ISO. \$\endgroup\$
    – jrista
    Jan 24, 2012 at 1:44
  • \$\begingroup\$ Here is a forum post on Canon's various methods of achieving ISO settings, and the consequences of each. I am not sure I fully agree with Browning's analysis (i.e. I would assume "Smart" ISO, which is simply storing the ISO boost amount in metadata for application in post, WOULD NOT be beneficial, but he claims otherwise, indicating "Bargain" ISO is good but flawed and all ISO should be done like "Smart" ISO.) \$\endgroup\$
    – jrista
    Jan 24, 2012 at 1:49
2
\$\begingroup\$

dynamic range is also loat with higher iso.

noise should be random, but some cameras exhibit banding - which can be limited to a single colour channel.

\$\endgroup\$
1
\$\begingroup\$

From a comment above clarifying the question: I meant to ask whether the amount of grain in the same photograph at high ISO differs for various colors.

In that case, the answer with current technology is:

Yes.

  • The blue channel is in general the noisiest.
  • The red channel is next.
  • And the green channel exhibits the least noise.

For details on why this is, see the answers to Why is the blue channel the noisiest?.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.