Is ISO the only determinant of noise?

[Andrew Taylor202717]

The world of digital photography is relatively new to me so forgive my lack of knowledge. Today was an unusually beautiful day in this neck of the woods, so I went on an orgy of picture taking.

I took a self-portrait sitting by the available light of a window inside my house at ISO 100, 1/80 shutter speed and 2.8 aperature. I expected that there would be almost no noise when I printed a 5 by 7, but there was quite a bit. I should mention that I chose to shoot in black and white. The picture looks good but there is a surprising amount of noise in it. Later in the day, I shot a picture of a river at ISO 50 and there's no sign of noise. Is there some reason why my ISO 100 black and white 2.8 picture had alot of grain?

 

[onions]

i did a few comparison shots a little while ago with different ISO settings and found the differences you would normally find between auto/50/100/200/400 (this was with an A80 btw). i also shot a few with a different white balance setting and found that the white balance setting can increase the noise levels quite a bit.

The world of digital photography is relatively new to me so forgive
my lack of knowledge. Today was an unusually beautiful day in this
neck of the woods, so I went on an orgy of picture taking.

I took a self-portrait sitting by the available light of a window
inside my house at ISO 100, 1/80 shutter speed and 2.8 aperature.
I expected that there would be almost no noise when I printed a 5
by 7, but there was quite a bit. I should mention that I chose to
shoot in black and white. The picture looks good but there is a
surprising amount of noise in it. Later in the day, I shot a
picture of a river at ISO 50 and there's no sign of noise. Is
there some reason why my ISO 100 black and white 2.8 picture had
alot of grain?

 

[Isabel Cutler]

Sometimes I will use a minus ev setting so that I can get a shutter speed fast enough to keep a handheld shot from blurring. Since I have to use an editing program to bring out shadow detail when I'm underexposing I will see lots more noise than if the picture were properly exposed.
Isabel

ase.com/isabel95
http://groups.yahoo.com/group/Digipets/
pBase supporter

 

[Tepa]

i did a few comparison shots a little while ago with different ISO
settings and found the differences you would normally find between
auto/50/100/200/400 (this was with an A80 btw). i also shot a few
with a different white balance setting and found that the white
balance setting can increase the noise levels quite a bit.

That is interesting... What white balance settings did you find to increase the noise? or what did you observe to have the least noise?

But to the original question. Resolution setting matters as well. All this may be clear to you already, sorry if this is too trivial ... The sensor in the camera has some physical dimensions, and some certain pixel number. If the camera is used set at its full resolution each pixel in the resulting image pretty much originates from a certain pixel in the sensor. However, if you choose a lower resolution setting in camera, each image pixel is an average of number of pixels in sensor and thus the noise is averaged out.

This was discussed earlier (few weeks ago?) in this forum, but I didn't find the posts of that discussion in my brief (10 sec) search

-tepa

 

[onions]

it was during the daytime and i took pictures of my desk using the natural ambient light. Auto white balance setting produced the least noise and just for kicks i tried the tungsten setting and that was what produced more noticeable noise.

i did a few comparison shots a little while ago with different ISO
settings and found the differences you would normally find between
auto/50/100/200/400 (this was with an A80 btw). i also shot a few
with a different white balance setting and found that the white
balance setting can increase the noise levels quite a bit.

That is interesting... What white balance settings did you find to
increase the noise? or what did you observe to have the least noise?

But to the original question. Resolution setting matters as well.
All this may be clear to you already, sorry if this is too trivial
... The sensor in the camera has some physical dimensions, and some
certain pixel number. If the camera is used set at its full
resolution each pixel in the resulting image pretty much originates
from a certain pixel in the sensor. However, if you choose a lower
resolution setting in camera, each image pixel is an average of
number of pixels in sensor and thus the noise is averaged out.

This was discussed earlier (few weeks ago?) in this forum, but I
didn't find the posts of that discussion in my brief (10 sec)
search

-tepa

 

[ANAYV]

According to Phil,the S50 and G5 are low in noise only at ISO50.Above that,and noise becomes an issue.

Even in my S30(3.2MP),if I stay on ISO 50(most all the time),noise is never an issue.But if I go up the ISO ladder,noise creeps in.

I recommend staying on ISO50 as much as you can,for the cleanest pictures.

ANAYV

i did a few comparison shots a little while ago with different ISO
settings and found the differences you would normally find between
auto/50/100/200/400 (this was with an A80 btw). i also shot a few
with a different white balance setting and found that the white
balance setting can increase the noise levels quite a bit.

That is interesting... What white balance settings did you find to
increase the noise? or what did you observe to have the least noise?

But to the original question. Resolution setting matters as well.
All this may be clear to you already, sorry if this is too trivial
... The sensor in the camera has some physical dimensions, and some
certain pixel number. If the camera is used set at its full
resolution each pixel in the resulting image pretty much originates
from a certain pixel in the sensor. However, if you choose a lower
resolution setting in camera, each image pixel is an average of
number of pixels in sensor and thus the noise is averaged out.

This was discussed earlier (few weeks ago?) in this forum, but I
didn't find the posts of that discussion in my brief (10 sec)
search

-tepa

 

[KEITH-C]

it was during the daytime and i took pictures of my desk using the
natural ambient light. Auto white balance setting produced the
least noise and just for kicks i tried the tungsten setting and
that was what produced more noticeable noise.

Tungsten setting is anyway inappropriate in ambient lighting in daylight or are you calling interior light ambient lighting?

Ps B&W pics with all the noise removed look odd to me anyway but I am perhaps used to older fim B&W when grain could be used to good effect.
--
Keith-C

 

[Ron Parr]

There are many factors that influence noise aside from ISO, including temperature and exposure time. Different colors also tend to exhibit different amounts of noise.

The effect of overall exposure on noise is often misunderstood. Increasing exposure, in general, does not decrease noise and actually increases noise. However, it reduces your percpeption of noise because brighter areas will have a higher signal to noise ratio.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Forbes]

me mystified, as I had been noticing as Isabel had mentioned that underexposrues caused much higher noise (this really makes sense, because in correcting the underexposure in editing we're "pushing" the film speed to a higher number, sometimes several stops, for select portions of the image), but I had not observed that overexposrue was also a concern for noise. At any rate, the exposure has everything to do with this. Don F.

There are many factors that influence noise aside from ISO,
including temperature and exposure time. Different colors also
tend to exhibit different amounts of noise.

The effect of overall exposure on noise is often misunderstood.
Increasing exposure, in general, does not decrease noise and
actually increases noise. However, it reduces your percpeption of
noise because brighter areas will have a higher signal to noise
ratio.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

--
http://www.DForbesRowanPhotos.OrangeCountyandSurrounding.PhotoShare.co.nz
A80: It's a neat camera, even if it's not black.
D. F. R.

 

[Ron Parr]

me mystified, as I had been noticing as Isabel had mentioned that
underexposrues caused much higher noise (this really makes sense,
because in correcting the underexposure in editing we're "pushing"
the film speed to a higher number, sometimes several stops, for
select portions of the image), but I had not observed that
overexposrue was also a concern for noise. At any rate, the
exposure has everything to do with this. Don F.

Well, please don't misunderstand - overexposure won't make your pictures look noisier.

Here's one way to think of it (note that these are just examples and not actual noise quantities): Suppose your images have brightness levels ranging from 0 to 255 (8 bits). Some noise sources will increase at a rate proportional to the square root of the brightness, so if you have a pixel with brightness value 25, + - 5 of this might noise. On the other hand, if you have an image with brightness value 225, + - 15 of this might be noise.

At first, the 15 may seem worse, but we need to remember that it's a much smaller percentage. 15/225 = 6%, 5/25 = 20%. So, the end result will be that the dark areas with brightness 25 will look noisier than the bright areas even though the bright areas have more noise in total.

So, getting nice, well exposed images is still a good way to reduce the appearance of noise in your photos.

Of course, this is a dramatic oversimplification with made up numbers. There are many other sources of noise, including errors from the A/D converter that also come into play. For example, it's possible that very dark areas will be subject to greater A/D conversion errors.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Forbes]

At first, the 15 may seem worse, but we need to remember that it's
a much smaller percentage. 15/225 = 6%, 5/25 = 20%. So, the end
result will be that the dark areas with brightness 25 will look
noisier than the bright areas even though the bright areas have
more noise in total.

So, getting nice, well exposed images is still a good way to reduce
the appearance of noise in your photos.

Of course, this is a dramatic oversimplification with made up
numbers. There are many other sources of noise, including errors
from the A/D converter that also come into play. For example, it's
possible that very dark areas will be subject to greater A/D
conversion errors.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

--
http://www.DForbesRowanPhotos.OrangeCountyandSurrounding.PhotoShare.co.nz
A80: It's a neat camera, even if it's not black.
D. F. R.

 

[Tepa]

The effect of overall exposure on noise is often misunderstood.
Increasing exposure, in general, does not decrease noise and
actually increases noise. However, it reduces your percpeption of
noise because brighter areas will have a higher signal to noise
ratio.

Thanks Ron for an excellent post.

I think that noise seen in higher ISO shots has also something to do with what Ron described above. But to be more specific, I think that the ISO setting is actually a GAIN setting. Otherwise it would be impossible to get the same exposure with i.e. 1/60,F2 @ISO 50 and with 1/500,F2 @ISO400. Any why do I think this? Because the sensor is the same. I don't think that the sensitivity of the sensor can be increased per se. Correct me if I'm wrong.

And what all this got to do with what Ron said in his earlier post? If ISO setting is actully a GAIN setting, by incresing gain in image we will end up increasing the electric noise in it as well, and thus signal-to-noise ratio decreases.

What do you think?

-tepa

 

[Elia Vecellio]

Hi,

I just wanted to add that I quite like the colours that I get if I do this in PSP8 (increasing the Gamma value and setting the mid-tone compression to -10 or so). I often set the Flash Exposure Compensation to -1 or so for night-time shots so that they come out underexposed and then use the same technique to get a more noir-ish sepia-ish/grunge-ish (for lack of a better word I can think of).

Elia

Sometimes I will use a minus ev setting so that I can get a shutter
speed fast enough to keep a handheld shot from blurring. Since I
have to use an editing program to bring out shadow detail when I'm
underexposing I will see lots more noise than if the picture were
properly exposed.
Isabel

ase.com/isabel95
http://groups.yahoo.com/group/Digipets/
pBase supporter

 

[Chris Eley]

And what all this got to do with what Ron said in his earlier post?
If ISO setting is actully a GAIN setting, by incresing gain in
image we will end up increasing the electric noise in it as well,
and thus signal-to-noise ratio decreases.

What do you think?

Absolute temperature is a major influence on the noise level
in all electronic systems (virtually all astronomical CCDs and
detectors are actively cooled to decrease noise...)

If you can keep the camera cool (but not so cold as to risk
causing condensation), you will see less noise in your pictures
at any ISO setting. But it's still best to keep the latter as low
as possible, and use a tripod/support if necessary.

HTH,

Chris

--
Chris Eley (Oxfordshire, UK)

 

[Ron Parr]

And what all this got to do with what Ron said in his earlier post?
If ISO setting is actully a GAIN setting, by incresing gain in
image we will end up increasing the electric noise in it as well,
and thus signal-to-noise ratio decreases.

I think we agree on this. Let me try to restate it and see if this sounds right to you:

When you boost the ISO, you are actually underexposing the image and then amplifying the output from the sensor to compensate. The underexposed image will have lower SNR.

A perfect amplifier would boost the signal without introducing any new noise, so the boosted output would have the same SNR as an underexposed image. However, amplifiers aren't perfect, so high ISO images can suffer from low SNR for the image that comes off the sensor, plus amplifier noise as it is boosted up to the right brightness level.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[bac522]

Since digital cameras don't really have film's ISO speeds, the ISO setting in a digital camera really refers to boosting or lowering the CCD's sensitivity to light (IE Gain). The problem with gain is that as the sensitivity level is raised, it produces noise (unwanted electronic artifacts in the picture, such as random white splotches or streaks), just as turning up the volume on a radio increases the amount of static you can hear in the background. Therefore, digital camera users should view gain in the same way that photographers look at ISO speed: turn it up when you need faster speeds or must shoot in low light, but be prepared for some image degradation.

BC

The effect of overall exposure on noise is often misunderstood.
Increasing exposure, in general, does not decrease noise and
actually increases noise. However, it reduces your percpeption of
noise because brighter areas will have a higher signal to noise
ratio.

Thanks Ron for an excellent post.

I think that noise seen in higher ISO shots has also something to
do with what Ron described above. But to be more specific, I think
that the ISO setting is actually a GAIN setting. Otherwise it would
be impossible to get the same exposure with i.e. 1/60,F2 @ISO 50
and with 1/500,F2 @ISO400. Any why do I think this? Because the
sensor is the same. I don't think that the sensitivity of the
sensor can be increased per se. Correct me if I'm wrong.

And what all this got to do with what Ron said in his earlier post?
If ISO setting is actully a GAIN setting, by incresing gain in
image we will end up increasing the electric noise in it as well,
and thus signal-to-noise ratio decreases.

What do you think?

-tepa

 

[Tepa]

I think we agree on this. Let me try to restate it and see if this
sounds right to you:

When you boost the ISO, you are actually underexposing the image
and then amplifying the output from the sensor to compensate. The
underexposed image will have lower SNR.

yep, that is exactly what I was (trying) to say. Thanks Ron.

A perfect amplifier would boost the signal without introducing any
new noise, so the boosted output would have the same SNR as an
underexposed image. However, amplifiers aren't perfect, so high
ISO images can suffer from low SNR for the image that comes off the
sensor, plus amplifier noise as it is boosted up to the right
brightness level.

right again.

So I'm not surprised to see increased noise when iso is increased, as amplifying lower SNR images certainly amplifies the noise in them as well causing greater variation in the resulting image. And then there is also this noise introduced by amplifiers... I personally use iso50 always when it is possible, something I would urge everybody to do, if visually noise-free images are preferred.

-tepa

 

[ANAYV]

Why don't the noise creep up as fast with Bigger pixels?

On the 10D,raising ISO levels increases noise,but at a much smaller rate.I'd guess the same amount of amplification is being applied,but the increase in noise is WAY smaller.

ANAYV

I think we agree on this. Let me try to restate it and see if this
sounds right to you:

When you boost the ISO, you are actually underexposing the image
and then amplifying the output from the sensor to compensate. The
underexposed image will have lower SNR.

yep, that is exactly what I was (trying) to say. Thanks Ron.

A perfect amplifier would boost the signal without introducing any
new noise, so the boosted output would have the same SNR as an
underexposed image. However, amplifiers aren't perfect, so high
ISO images can suffer from low SNR for the image that comes off the
sensor, plus amplifier noise as it is boosted up to the right
brightness level.

right again.

So I'm not surprised to see increased noise when iso is increased,
as amplifying lower SNR images certainly amplifies the noise in
them as well causing greater variation in the resulting image. And
then there is also this noise introduced by amplifiers... I
personally use iso50 always when it is possible, something I would
urge everybody to do, if visually noise-free images are preferred.

-tepa

 

[Ron Parr]

Why don't the noise creep up as fast with Bigger pixels?

On the 10D,raising ISO levels increases noise,but at a much smaller
rate.I'd guess the same amount of amplification is being
applied,but the increase in noise is WAY smaller.

Large pixels collect more photons and can hold more electrons. Assuming the input range on the A/D converter is the same, then larger pixels would need less amplification, but this is just part of the story.

The other part of the story is that the larger signal typically means a higher signal to noise ratio for the reasons given above. (Note the following numbers are made up to make the example simple.) Let's say a small pixel saturates at 1000 electrons and a pixel with 5X the area saturates at 5000 electrons, so these values determine what it means for each of these sensors to detect pure white. On the larger sensor, the signal has increased by 5X, but the noise from dark current and photon shot noise have increased by only the square root of 5, so the signal to noise ratio of the larger sensor is increased.

Of course, there are other sources of noise and they don't all scale this way, but this is a nice example that shows how using larger pixels can improve SNR - at least with respect to some noise sources.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[ANAYV]

Thanks for that info.

Why don't the noise creep up as fast with Bigger pixels?

On the 10D,raising ISO levels increases noise,but at a much smaller
rate.I'd guess the same amount of amplification is being
applied,but the increase in noise is WAY smaller.

Large pixels collect more photons and can hold more electrons.
Assuming the input range on the A/D converter is the same, then
larger pixels would need less amplification, but this is just part
of the story.

The other part of the story is that the larger signal typically
means a higher signal to noise ratio for the reasons given above.
(Note the following numbers are made up to make the example
simple.) Let's say a small pixel saturates at 1000 electrons and a
pixel with 5X the area saturates at 5000 electrons, so these values
determine what it means for each of these sensors to detect pure
white. On the larger sensor, the signal has increased by 5X, but
the noise from dark current and photon shot noise have increased by
only the square root of 5, so the signal to noise ratio of the
larger sensor is increased.

Of course, there are other sources of noise and they don't all
scale this way, but this is a nice example that shows how using
larger pixels can improve SNR - at least with respect to some noise
sources.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/