High ISO and noise with correct exposure

[Alis]

I have a question for the experts here: If the exposure is correct (at the pixel level), would a high ISO setting still be associated with high noise?

Case: You are using flash to shoot a scene (indoors). The subject is under a recessed light. When you adjust for the foreground, the background looks underexposed and dark. You increase the ISO to compensate for that. Now if the ISO setting is increased here, would the correctly exposed areas of the image be noisy?

I hope I have stated the question clearly.

Thanks!

[Manfred M]

Ali, noise in an image is not directly related to correct exposure, but rather high ISO levels, and this is totally related to the way ISO levels are created in a digital camera. A sensor only has a single sensitivity, and all usable output is due to the amplification of the signal from the sensor. Low ISO has a relatively low level of amplification, while a high level of amplification results in a high ISO setting. The technical term for the level of amplification is gain.

In the sensor and amplifier circuitry, there is the data from the conversion of photons to electrons (signal), and extraneous noise from other sources like the electronic circuits themselves. Unfortunately, the amplifier cannot distinguish between the signal and the noise and amplifies both. This is referred to the signal to noise ratio (S/N). There are relatively few photons associated with darker values in the image, so the noise component tends to make up a higher portion of the total signal, which is why noise is usually more visible in the dark areas of the image.

Now to your example, it’s hard to guess at outcomes without knowing exactly what you are shooting, the ISO you are shooting at, noise characteristics of your camera etc. I assume that you are using two light sources; the overhead recessed light plus your flash. It seems to me that that the low ISO setting should give you a nice shot, with good separation of the subject from the background. With a relatively low ISO value, the noise in the background should be quite low. If you now increase the ISO setting to brighten up the background, and you are illuminating the subject properly (hopefully the extra light on the subject does not blow out the highlights), you are still going to have a fairly dark background and the noise level might be more obvious than with a lower ISO setting.

Again, these are all guesses. What I suggest you do is shoot both ways and see which comes out looking better. You might be able to do something in PP as well.

[Colin Southern]

Hi Ali,

What Manfred said, but it's really quite a complex issue with a number of variables.

Higher ISO modes use less and less of the sensor's potential - but - whether or not noise is visible depends on the ISO used for a given camera (or more specifically the signal-to-noise ratio, which varies between cameras), where the image is exposed within the range of possible exposures for that ISO setting (which also affects the signal-to-noise ratio), and to what degree one is magnifying the image to look at it, and the dynamic range of the scene being shot.

Without getting too complex - basically - normal camera metering usually leaves a 1.3 to 2 stop "safety margin" in a RAW exposure (ie an area that highlights can be exposed into if needed). At low ISO settings where the dynamic range capability of the camera is huge - there is still a huge margin between shadow detail and sensor noise -- so the safety margin doesn't make things visibly worse (thus ETTR doesn't gain you anything in most cases). At higher ISO modes though, the effective sensor dynamic range capability is reduced - and at very high ISO modes having a 2 stop safety margin (or two stops of potential highlight storage space wasted) pushes everything else closer and closer to the noise floor - at which point it may well become a problem. Put another way, at high ISO settings it becomes more and more important to push exposures to the limits of clipping to minimise noise.

The other issue is just how visible is the noise when looking at the shot at a whole? If one is going to pixel-peep then it'll look bloody aweful - but - when looking at the entire shot chances are it'll look just fine. So normally my advice is to just not worry too much about it.

So to answer the question - Yes, an "correct" exposure (in the sense of "looking correct on the camera") will give you more noise at high ISO settings than an "incorrect" exposure that looks over-exposed (but is not blown).

Does that make sense?

In real-world terms, it's often a compromise; in your case you can only open the lens to a certain point before you risk DoF issues, and you can only lower the shutter speed to a certain point before you risk camera shake or subject motion ruining the portion of the shot lit by ambient light. After that, ISO is the only other variable. In practice, crank the ISO up (for your chosen shutter speed and aperture), and if it's going to be a very high ISO (say 6400 or above) then give some serious thought to adding some +EC (but just ensure you don't blow the highlights in critical areas).

High ISO is a bit like heart care in the elderly in that the margin for error gets less and less the higher the numbers go.

[darkslide]

Here's an example - a shot I took on Saturday. I can see a little noise when the image is enlarged on my screen, but nothing intrusive in the part that is correctly exposed - i.e.; his face.

125s f/2.8 ISO 8000

[Colin Southern]

Here's an example - a shot I took on Saturday. I can see a little noise when the image is enlarged on my screen, but nothing intrusive in the part that is correctly exposed - i.e.; his face.

125s f/2.8 ISO 8000

Perfect example of noise "not being an issue" - what the chap REALLY needs is a haircut though!

[darkslide]

I don't choose the groups that play here! (He's Australian....plays in a group called The Hard Ons...)

[dje]

I don't choose the groups that play here! (He's Australian....plays in a group called The Hard Ons...)

We Aussies are so refined aren't we !

(Not quite as refined as the Kiwis though....)

[Colin Southern]

I don't choose the groups that play here! (He's Australian....plays in a group called The Hard Ons...)

He needs to "get a haircut and get a real job"

[ajohnw]

Another way of looking at this aspect is how many dark areas there are in the shot and just how dark they are. Black as far as a sensor is concerned is zero, no signal at all so any noise is likely to be apparent. On the other hand at the other end of the scale a bright colour that gives a signal that is the maximum the sensor can record tends to mask any noise as it's no longer as significant. In practice it isn't as simple as that as a camera might make a more decent job of pitch black no light at all and get worse as light levels increase and then eventually improve as the noise gets less significant in relation to the light levels that are being recorded.

The amount of noise a camera has at various iso settings varies with make so there isn't a hard and fast rule for how high an iso can be used to take a shot. It depends on the camera and the scene and in the end just how effective noise reduction software can be. If high iso's are used it may pay to try using jpg's once the camera is at a level or situation where it struggles. Auto too as camera manufacturers go to great lengths to try and get the best possible results under these circumstances. There is also specialist software available. Pass on the name.

An example of high iso causes problems is this one. 1600 iso on an Olympus Pen. Rather a lot for that camera but it copes. Many cameras will offer higher iso's. This shot is straight out of the camera. Noise reduction has lost detail in the blacks and if you right click open image with any luck you will see the full sized image and signs of noise all over it if you click on it again except in the blacks. The camera white balance has also shifted, That's not unusual either, The E-PL1 is crippled in software to 1600 max but going on this shot I suspect the camera would struggle if set higher and if I did want detail in blacks I don't think that the camera could record them on a shot like this. If I adjusted the exposure noise would become more apparent. The white balance shift is odd if you look closely. Shirt colours are correct.



-

[Shadowman]

Here's an example - a shot I took on Saturday. I can see a little noise when the image is enlarged on my screen, but nothing intrusive in the part that is correctly exposed - i.e.; his face.

125s f/2.8 ISO 8000

If that was shot with flash, wouldn't there be light fall off at a certain distance which would result in high noise areas ?

[darkslide]

I don't use flash - ever.

However I find it interesting that in the out-of-focus areas (his right arm, the lower part of the mic stand etc.) that the noise is much more "apparent" and I'm wondering if this a real artifact or simply that we don't "see" the noise in the well defined, sharp, areas?

[MrB]

The other issue is just how visible is the noise when looking at the shot at a whole? If one is going to pixel-peep then it'll look bloody aweful - but - when looking at the entire shot chances are it'll look just fine. So normally my advice is to just not worry too much about it.

Ian, your photo of the guitarist seems to be a good example of what Colin often tells us. However, you write that you can see noise when you enlarge the image on your screen. The CiC tutorial defines it like this - "...noise appears as random speckles on an otherwise smooth surface..." - when your image is at 100% on my screen, this effect is not apparent to me (a relatively inexperienced photographer). How deeply do you need to pixel-peep, or perhaps this is just my old eyesight, or my spectacles, or my monitor, or perhaps I still don't understand what to look for?

Philip

[darkslide]

No, I think you're correct - it's me nit-picking. It's just out-of-focus rather than noise! (If you look at the guitarists right arm it gradually degenerates - this is simply focus rather than noise. My apologies for creating the confusion!!)

[ajohnw]

I don't use flash - ever.

However I find it interesting that in the out-of-focus areas (his right arm, the lower part of the mic stand etc.) that the noise is much more "apparent" and I'm wondering if this a real artifact or simply that we don't "see" the noise in the well defined, sharp, areas?

I should have added that clever noise reduction software may have problems with parts of a shot that isn't in focus. There seems to be a weird effect on the red by the players right arm as well. I don't think all of that is down to movement. Simple noise reduction is a blur and some from of sharpening so anything that is already blurred may have problems. The other effect is light levels against noise levels in the sensor.

If you look at the shot I posted full frame you will see a certain amount of fuzzyness in a number of places.The lens used is sharp the effect is down to noise reduction. If you look at other parts of the image slight mushy noise can be seen where the noise reduction software has smoothed it out.

-

[Mike Buckley]

We can say whatever we think about the musician's profession and haircut, but all photographers are surely jealous of his neck strap!

High ISO is a bit like heart care in the elderly in that the margin for error gets higher and higher the higher the numbers go.

That's a really great analogy that I have never thought of. However, I think you meant to say that the margin for error gets lower and lower, not higher and higher.

[Colin Southern]

We can say whatever we think about the musician's profession and haircut, but all photographers are surely jealous of his neck strap!



That's a really great analogy that I have never thought of. However, I think you meant to say that the margin for error gets lower and lower, not higher and higher.

Oops - quite right (fixed). Sorry Ali - looks like I failed cardiology 101

[darkslide]

We can say whatever we think about the musician's profession and haircut, but all photographers are surely jealous of his neck strap!

Help! Does no-one on this forum shoot NIKON?????

Ray, the bass player, got pretty active too - FENDER will no doubt be please - at least I got that bit in focus!



Strangely, this one, taken at ISO 5600 'seems' to have more visible noise...

[Alis]

Many thanks, guys! I was sleep while all this happened. So I am still digesting it. But From reading them once I am not sure if I know the answer to my question. It was sort of a theoretical question but if you want example here is what made me ask this question although I has always been in my mind (slow mind ):



So, my theoretical question is: If I am a single pixel and I am correctly exposed, do I generate a lot of noise? And here is why this is a question for me: The background is also correctly exposed but at the expense of a longer exposure time. The foreground however is not exposed for more than the duration of the flash. So, I thought, and as Colin mentions, since that is the only choice sometimes to increase the ISO, that advice is based on the fact that theoretically where you get good light in the foreground or at the pixels that capture those areas, the noise issue is not as much of a problem as the background, regardless of what you can do in PP or with noise reduction softwares.

Am I being dumb and you guys already answered my question or no?

Cheers!

[Alis]

Oops - quite right (fixed). Sorry Ali - looks like I failed cardiology 101

Yes, so to make up, add 2 miles to the daily jogging distance... And stick to photography

[ajohnw]

Many thanks, guys! I was sleep while all this happened. So I am still digesting it. But From reading them once I am not sure if I know the answer to my question. It was sort of a theoretical question but if you want example here is what made me ask this question although I has always been in my mind (slow mind ):



So, my theoretical question is: If I am a single pixel and I am correctly exposed, do I generate a lot of noise? And here is why this is a question for me: The background is also correctly exposed but at the expense of a longer exposure time. The foreground however is not exposed for more than the duration of the flash. So, I thought, and as Colin mentions, since that is the only choice sometimes to increase the ISO, that advice is based on the fact that theoretically where you get good light in the foreground or at the pixels that capture those areas, the noise issue is not as much of a problem as the background, regardless of what you can do in PP or with noise reduction softwares.

Am I being dumb and you guys already answered my question or no?

Cheers!

Add on to my other reply

The correctly exposed back ground is only likely to happen if you use fill in flash or a variation of it. Another way of putting that is that a correct exposure of the background in your shot might over expose the subject. What I suspect has happened is that the flash has been used to set the exposure of the subject and that the actual exposure time is insufficient to record the background correctly. It may have helped a bit but the main lighting has come from the flash - background is further away so there is less light intensity on it. If you have a problem understanding that googling "flash guide numbers" should help. These are used to set manual flash exposures purely based on distance and aperture. and of course iso.

-