[UserAccessDenied]

I'm still new to photography; I think I learn something new every time I shoot.

I've been shooting manual mode from the beginning. I disabled SR, ISO NR, Shutter Speed NR, etc.
I bought a few photography books when I started out, and still refer to them from time to time.

One thing I was told, from reading those books and hearing from other photographers, keeping ISO as low as possible is key to reducing noise - That "noise is directly related to ISO". Of course with a caveat - as long as you can keep ISO low enough and still get the shot. If you need a faster shutter speed or a smaller aperture, then ISO will have to compensate.

Well, I took that to heart and rarely ever took my ISO off 100. If I got to ISO 400 I often tried to deal with a slower shutter speed or opened one more F-stop to get the shot without increasing ISO anymore.
Turns out this is not the way to do it!

I took this shot the other day - nothing special, but I wanted to focus on why it is so noisy:
ISO 100 - f6.3 - 1/800 @ 300mm
on paper that sounds great, right?





Here is my theory:
ISO does not have a direct relationship with noise.

If there is a lack of light, there will be noise regardless of ISO setting.
If there is an abundance of light, there will be little noise; though the image may be overexposed due to the ISO setting.


Forcing the camera to ISO 100 in this case just underexposed the image. This resulted in heavy PP to get the image to where it is now.
If I allowed the camera to properly expose (let's just say ISO 800), then I'd probably attribute the noise to shooting at ISO 800.

If this is true, fixing aperture and shutter speed to a constant, ISO has no effect on the level of noise in an image (assuming each photo gets properly exposed in PP).


I'd really like to test this out further, I might do something this weekend if I find time...
Obviously the one photo above doesn't render scientific results. But it got me thinking and researching the topic a bit - It's fascinating to think I've been so naive the last two years!

Why do I shoot manual?
I want to control the depth of field and the degree of motion in my image.
I may not necessarily want control over the sensitivity of my camera's sensor.

I'll likely always set aperture and shutter speed manually, but now I think I'm okay with letting the K-3ii choose it's own ISO.


What are your thoughts?
If I'm wrong about this, please explain in the comments below.
Like I said, I learn something new everyday.

---

Last edited by UserAccessDenied; 12-13-2016 at 04:35 PM.

[UncleVanya]

This is a very complex subject. ISO performance varies by sensor. Some sensors are ISO invariant which means that when underexposed the effect is similar to cracking up ISO. Some sensors require the ISO to be turned up to get the sensor sensitivity high enough.

[Ontarian50]

Agree it's a complex subject.

Looking closely at your image, it appears the "noise" patterns are only apparent in the out of focus areas - the background, and even the animal's nose.

But where you have sharp, fine detail, noise is nowhere to be seen. This is very similar to film images, that show film grain in out of focus areas, but areas with fine detail show no grain.

In the digital world, it must have something to do with the image processing not having strong edges or contrast to quantify, and it ends up putting a random pattern over the softer areas.

Was this processed from raw, or is an in-camera jpeg. Of course, either way, JPEG compression gets a bite out of it.

[UserAccessDenied]

This is a very complex subject. ISO performance varies by sensor. Some sensors are ISO invariant which means that when underexposed the effect is similar to cracking up ISO. Some sensors require the ISO to be turned up to get the sensor sensitivity high enough.

Does this mean camera to camera or between brands and models?
So my K-3II could be more sensitive to light than another K-3II?

---------- Post added 12-13-16 at 02:41 PM ----------

Agree it's a complex subject.

Looking closely at your image, it appears the "noise" patterns are only apparent in the out of focus areas - the background, and even the animal's nose.

But where you have sharp, fine detail, noise is nowhere to be seen. This is very similar to film images, that show film grain in out of focus areas, but areas with fine detail show no grain.

In the digital world, it must have something to do with the image processing not having strong edges or contrast to quantify, and it ends up putting a random pattern over the softer areas.

Was this processed from raw, or is an in-camera jpeg. Of course, either way, JPEG compression gets a bite out of it.

Shot DNG - brought up the exposure in LR and then exported to JPG.

I know it's very complex.
If anything I'm just glad to realize I shouldn't worry too much about ISO anymore.

If there is a lack of lighting, I should still shoot for the aperture and shutter speed I want rather than compensating for the lower ISO.

[ChristianRock]

3 steps above ISO 100 = ISO 800.

If you shoot RAW at ISO 100 and then you have to bring up exposure 3 steps in PP, you get similar amounts of noise as if you had shot at ISO 800 - or even worse if you're shooting at 12-bit.

Disclaimer: This is not a scientific statement, it's more of what I have learned in practice. Reality may vary, but I don't think I'm too far off the mark here My observation is that dynamic range and ISO performance sort of go hand in hand.

[cali92rs]

Also remember that dark areas carry more noise than bright areas...this is the reason for "expose to the right".

[biz-engineer]

Shooting low ISO underexposed can never improve noise, and shooting higher ISO overexposed can eventually reduce noise whenever ISO setting isn't purely done in software. Whenever on sensor analog to digital conversion range is adjustable, noise is slightly lower when using higher ISO, for the same given light , aperture and shutter speed. Downsizing images can also reduce or worsen the appearance of noise in OOF areas, when the noise isn't fully random and interpolation re-sample of the same image that may induce some form of aliasing, noise comes in clusters and become more visible, the solution in this case is to resize to a slightly different size in order to avoid noise to come in clusters, IMO.

[UncleVanya]

My comments were about sensor model to model differences.

[normhead]

Yet the you look down the images at place like Imagine Resources where they show the same image at different ISOs, you can see the progression as noise gets worse and worse. I'm convinced part of this is the number of gradations to the right and left of your histogram. To the left right of the histogram , you have thousands of gradations, to the left hundreds. Your contrast is higher, but the noise is more visible. You can get noise in almost any image by pushing a already narrow histogram into the let half of the range when exposing then using levels to extend it to the full output range. You get great contrast but more noise. Using the right side you get less contrast, and less noise.

Having a nicely balanced histogram really help avoid noise while still giving you some contrast and saturation.

When we get a particularly noisey image, we go through a protocol. Turn of default contrast, sharpening and edge sharpening, definition, mid tone contast.and anything else that might emphasize the noise. It's rare we actually have to go to the denoise slider, which can seriously cost you resolution.

[photoptimist]

Shot DNG - brought up the exposure in LR and then exported to JPG.

I know it's very complex.
If anything I'm just glad to realize I shouldn't worry too much about ISO anymore.

If there is a lack of lighting, I should still shoot for the aperture and shutter speed I want rather than compensating for the lower ISO.

You still need to worry about ISO.

Whether you brighten the picture by increasing the ISO on the camera or do so in post-processing, you're going to be getting more noise in the image.

In available light photography, the photographer must often decide whether to accept more noise, more motion blur, or a shallower depth-of-field (and perhaps more image softness associated with a wide-open lens).

The "best picture" might not use the best aperture and shutter speed if it means using a worse ISO.

[TonyW]

As you have learned keeping ISO low i.e. at base will certainly help towards keeping the noise signature down to a minimum.

But the theory is incorrect:

ISO does NOT have a direct relationship with noise.

ISO DOES have a direct relationship with noise. You could think of your sensor having a fixed ISO, the so called base ISO and also the fact that the sensor has a noise signature.

This signature is pretty much the same at all ISO but to gain the effect of an increase in ISO the camera electronics amplify the signal and in doing so also amplify the noise as well.

At base ISO the signal to noise ratio is as good as it gets with little evidence of noise, but as we raise ISO the camera electronics must amplify the signal and eventually we reach the point at high ISO where we have an unacceptably low signal to noise ratio. This is dependent on image intentions and your tolerance for noise

If there is a lack of light, there will be noise regardless of ISO setting.

Well, yes and no, but I suspect you may be thinking about this in the wrong way.

First as already covered there will be the same noise signature at ALL ISO some more acceptable than other.

Lack of light i.e. photons hitting the sensor needs to be controlled by exposure time and aperture. So that you are in control of the amount of light that falls on the sensor.

If there is an abundance of light, there will be little noise; though the image may be overexposed due to the ISO setting.

If you do not expose correctly regardless of the amount of light you will affect the noise signature. It is quite easy to underexpose with an abundance of light if relying on the camera meter e.g. in a snow or sand scene come to mind.

Overexposure is not caused by ISO settings but by poor choice of s/s and aperture for the subject

Forcing the camera to ISO 100 in this case just underexposed the image. This resulted in heavy PP to get the image to where it is now.

No, your choice of exposure for the ISO 100 setting was incorrect leading to the underexposure.

Your PP to bring the image to normal is the cause of the accentuated noise.

If I allowed the camera to properly expose (let's just say ISO 800), then I'd probably attribute the noise to shooting at ISO 800.

No what you have done by underexposing at ISO 100 and then bringing up in post is similar to what happens within camera.
The difference being that the camera ISO 800 setting boosts the signal of the sensor base (ISO 100?) to arrive at a lighter rendered image whereas you are attempting to do the same with your pp.

You are likely to find that the noise characteristics between an image shot at ISO 800 in camera vs shooting -3 EV (=ISO 100) and bringing up in post are very similar, but you would need to test your system

[mohb]

Slightly off topic, why did you choose to turn off SR? If you want to stay at a low ISO SR is going to give you a lot more latitude in choosing your other settings.

[Tas]

I took this shot the other day - nothing special, but I wanted to focus on why it is so noisy

There's plenty of better qualified members already chiming in on the camera ISO setting so I will stay out of the primary part of your post and raise another touched on by TonyW.

Post Processing will have an impact on the the noise in your final image, as will how it will be displayed; electronically or in print. The final format is a different rabbit hole, so I'll leave that there.

I would like to suggest if you're not already doing so is to only sharpen the elements you want sharp in an image. This means all the noise in the out of focus background of your image will not be enhanced with extra contrast. LR applies a default sharpening on import unless you change the settings so the first time you see your image in LR on a larger screen the noise will be enhanced.

Try pushing the sharpen slider left (to '0') then applying sharpening in your image using a brush bbut only to those areas you want sharpened. It's slower but for those images with higher ISO the noise should be less obtrusive, especially where the image is correctly exposed at capture.

Tas

[Rondec]

If you underexpose by three stops and brighten in Lightroom, you will have at least the same amount of noise as if you shot with iso 800 in the beginning. The only thing that may be a little better on the under exposed, low iso shot is dynamic range -- that is to say that there are times that I choose to underexpose to protect the highlights and then brighten the shadow areas in post, whereas if I shoot at higher iso and ignored the highlights, they would be unrecoverable.

[JinDesu]

Image noise is the sensor (of film, or music recorder, or whatever) lacking enough distinct signal to differentiate from general ambient signal - in the case of a image sensor that ambient signal is the general electrical field around the sensor. Where the incoming signal is not distinct, the local ambient signal may trigger the sensor pixels to register as red, green, or blue, and create the noise. Just as if you had a microphone and you hear static from the ambient electrical fields around the signal delivery circuits within.

ISO determines how much electrical amplification is applied to the sensor. A higher ISO is a higher electrical amplification to improve sensitivity to the lower input signal. Higher electrical amplification has the side effect of having a strong electrical field around that sensor, and therefore has a better chance of triggering the sensor with noise instead of the incoming signal.

Darker parts of a scene will have noise as the incoming signal from those areas may not be strong enough to differentiate from the electrical ambient signal. If you shoot a sunlit scene with a dark shadow even at ISO 100, the electrical ambient signal on the sensor may be stronger than the few photons you receive from that region in the shadow.

Similarly, out of focus regions may generate more noise because the incoming signal is seen as indistinct by the sensor.